Psyehopharmacologia (Berl.) 40, 33--52 (1974) 9 by Springer-Verlag 1974
Alcohol and Information Processing Van K. Tharp Jr., O. H. Rundell Jr., Boyd K. Lester, and Harold L. Williams Department of Psychiatry and Behavioral Sciences, University of Oklahoma Health Sciences Center, Oklahoma City, Oklahoma Received May 13, 1974; Final Version July 5, 1974 Abstract. Three experiments are reported which investigate the effects of acute alcohol intoxication (average blood alcohol concentration 100 rag-~ on some aspects of human information processing. The results are interpreted within the framework of a general information processing model (Smith, 1968), using the Sternberg (1969b) additive-factor method of analysis. Alcohol consistently impaired information outputting operations (i.e., response selection-organization), rather than information inputting operations (i. e., stimulus preprocessing and encoding). _Key words: Alcohol -- Information Processing -- Response Selection -- Additive Factor Analysis -- Reaction Time.
Most published reports of the effects of alcohol on h u m a n performance have been empirical and task specific rather t h a n theoretical in conception. Thus alcohol causes impairment on such tasks as time estimation (I~utschman and Rubenstein, 1966), reaction time (Moskowitz and l~oth, 1971), verbal retention (Jones, 1973) and character recognition (Huntley, 1972). Although all of these tasks require information processing, the results of such investigations do not reveal which processing operations m a y be most vulnerable to alcohol. One way to address this question is to examine alcohol effects from the perspective of an information processing model. The principal aim of the three experiments reported here was to test the utility of a serial information processing model for the understanding of certain performance deficits found in intoxicated h u m a n subjects (Ss). For character-recognition tasks, several models of information processing postulate a sequence of distinct processes or stages which intervene between presentation of a stimulus and initiation of a response (e.g., Norman, 1970; Smith, 1968; Sternberg, 1969a, b). A typical sequence of such hypothetical stages would be 1. stimulus preprocessing at a sensory-perceptual level, followed b y 2. stimulus categorization, wherein the stimulus item or information about the item is compared to other items stored in memory, tbllowed b y 3. response selection and organization, and 4. response execution (see Smith, 1968, for a review of 8
PsYchopharmacologia (Berl.), Vol. 40
34
V.K. Tharp Jr. et al.
this and other models of information processing). Such paradigms usually contain the testable assumption that the successive stages of information processing are additive and non-overlapping. Suppose that reaction time (I~T) is the dependent variable. Then an additive model assumes that the total time between stimulus and response is divided into discrete, non-overlapping intervals, each of which corresponds to one of the serial stages listed above. Obviously, two different experimental treatments could cause the same quantitative impairment of RT by influencing entirely different cognitive processes. Moreover, a drug such as alcohol might cause slowing in any or all of the stages described above ; it might cause overlap between normally discrete stages; it might alter the normal sequences of cognitive stages ; or it might reduce the accuracy of performance by impairing the output of one or more stages. I f the concept of distinct, additive cognitive stages is valid, then one should be able to identify specific experimental treatments that influence differentially and selectively each of the processing stages in the model. Sternberg's (1969a, b) approach to this problem is to conduct multipletreatment experiments, examining patterns of additivity and interaction between the effects of rationally selected experimental variables. The concept of distinct additive stages of information processing implies that the effects of experimental treatments which selectively influence different processing stages should be statistically independent. That is, the absence of significant positive interactions 1 between rationally selected treatments would support the hypothesis that these treatments influence different stages of information processing. Conversely, two experimental conditions selected a p r i e r i , whose effects show significant positive interactions, probably influence the same information processing stage. Using this "additive factor" method to examine performance on characterrecognition tasks, Sternberg (1967, 1969a, b) identified several experimental conditions, each of which apparently influences a specific stage of information processing. For example, his results indicate that a treatment which alters the discriminability of the stimulus probably influences the stimulus-preprocessing stage, whereas a treatment that alters either the difficulty of mapping the stimulus on the response or the distribution of response probabilities probably influence the stage of response selection and organization. Hereafter in this paper, a task-related treatment whose probable locus of effect has already been identified will be called an "established treatment". 1 A positive interaction between two treatments is a relationship in which their joint effect is greater than the sum of their individual effects. The relationship between two treatments may be described as additive when their joint effect is equal to the sum of their individual effects
Alcohol and Information Processing
35
These ideas suggest a line of investigation which may reveal the locus or loci of a drug effect on human performance. Thus, patterns of additivity and interaction between the effects of alcohol and those of established treatments could indicate which cognitive process or processes (i.e., processing stages) are most vulnerable to the drug. For example, a strong positive interaction between the effects of alcohol and those of an established treatment would imply that both experimental variables influence the same stage of information processing. On the other hand, an additive (statistically independent) relationship between the effects of alcohol and an established treatment would suggest that each experimental variable influences a separate processing stage. A few recent studies have already examined certain effects of alcohol ,on information processing. For example, experiments by l~oskowitz and his colleagues (Moskowitz and Burns, 1971; Moskowitz and DePry, 1968; Moskowitz and Roth, 1971) indicated that alcohol generally slows the rate at which information can be processed. These investigators have laot related their results to a specific information-processing model, but t t u n t l e y (1972), using an approach similar to Sternberg's, concluded that alcohol impaired more "central" processing stages. Thus, either stimulus categorization or response selection appeared to be more vulnerable to alcohol than "peripheral" stages such as stimulus preprocessing or response execution 2. The series of experiments reported here examined the combined effects of alcohol and several established task-related variables on three character-recognition tasks.
Experiment 1: Effects of Alcohol, Stimulus Discriminability, and Stimulus-Response Compatibility on Letter Recognition Accuracy Unpublished pilot studies in this laboratory had shown that alcohol impaired the recognition of tape-recorded letters. Within the sequential stage model, accurate transcription of these stimuli presumably depends on the integrity of at least two stages: stimulus preprocessing and response selection-organization. Experiment 1 used two task variables, stimulus discriminability (DISC) and stimulus-response compatibility (SRC), in an effort to assess the integrity of stimulus preprocessing and response selection-organization, respectively, under alcohol intoxication. Since the two task-related established treatments had previously shown additive effects on R T (Sternberg, 1967), we did not anticipate a significant positive interaction between their effects either in sober or intoxicated Ss. 2 The S~ernberg additive factor method has been used recently by Darley, Tinklenberg, Hollister, and Atkinson (1973) to investigate the effects of marijuana information processing. 3*
36
V . K . Tharp Jr. et al.
Method Subjects. The Ss were 18 male volunteers (ages 21--30 years) with normal hearing who were randomly assigned to alcohol and placebo groups of nine Ss each and tested individually. They were recruited from medical and graduate programs of the University of Oklahoma Health Sciences Center and were paid for their participation. In this and the other two experiments, all Ss were familiar with alcoholic beverages to the extent each had occasionally drunk moderately at evening social affairs. None acknowledged frequent use of other psychoactive drugs, and none were on prescription medication. Procedure. The stimuli consisted of tape-recorded lists of six letters randomly selected from the 21 consonants of the English alphabet, with the constraint that a given letter occurred only once in each set of six. The letters were presented through professional quality earphones over a background of white noise. Single letters were presented at a rate of one per second, with a 10 see rest period interposed between each set of six. The letters were presented at an average sound pressure level of about 83 decibels (dB, re 0.0002 dynes/em2). Stimulus discriminability (DISC) was varied by altering the loudness of background noise, and hence the signal-to-noise ratio. Thus, background noise was set at 70 dB for the high DISC condition, and at 80 dB for the low DISC condition. S's task in the one-second interval between stimuli was to print (in capital form) either the letter presented (high stimulnsresponse compatibility) or the letter next in succession in the alphabet (low SI~C). Following a practice session with 24 lists, S performed the task in each of the four combinations of the task variables in a balanced order. With a total of 240 six-letter lists, the first and last 60 lists were presented under the high DISC condition, whereas the middle 120 were presented with low DISC. The first, fourth, fifth, and eighth blocks of 30 lists each, were presented under high St~C and the remaining four blocks under the low compatibility condition. Accuracy of transcription was scored by two research assistants who were not familiar with the conditions of the experiment. Each S was fasted for 4 firs prior to beginning the experiment, and all Ss were tested individually at the same time each day. Ss in the alcohol group received 1.32 ml of 95~ ethanol per kg body weight. The ethanol was mixed 1:4 with ginger ale or orange drink, according to preference, and divided into three drinks, the S being given 10 rain to consume each drink. The practice session followed immediately upon consumption of the third drink. Placebo Ss received three drinks of ginger ale or orange drink with 4.0 ml of 95~ ethanol floated on top of each to produce the smell and taste of alcohol. An initial measure of blood alcohol concentration (BAC, Stephenson ~odel 900 Breathalyzer) was taken 15 rain after consumption of the final drink. The mean BAC for the alcohol group at this time was 100 mg-~ (range: 80--120 mg-~ Two additional BAC readings were taken at the middle and end of the approximately 1-hr experimental session. For individual Ss, the maximum variation in the 3 readings was about 20 mg-~ All BAC readings taken on placebo Ss were well below 10 mg-~ (at about the noise level of the Breathalyzer). The same ethanol dosage schedule and BAC measurement procedures were followed for all 3 experiments.
Results As can be seen in T a b l e 1, a v e r a g e error r a t e s (per 60 s i x - i t e m lists) inc r e a s e d in level 2 o f each e x p e r i m e n t a l condition. Since t h e v a r i a n c e s were n o t h o m o g e n e o u s across t r e a t m e n t c o m b i n a t i o n s , analysis o f v a r i a n c e was
Alcohol and Information Processing
37
Table 1. Effects of alcohol, stimulus discriminability, and stimulus-response compatibility on registration of letters a SRC condition
Placebo
Alcohol
High DISC
Low DISC
High DISC
Low DISC
HighSRC
X s
3.78 1.81
25.11 9.29
6.33 3.92
33.33 11.22
Low SRC
X s
8.56 3.07
32.56 12.13
18.22 6.55
53.33 16.05
a Total errors per 60 lists.
performed on the square-root transform of the raw scores. Results of this analysis were confirmed b y appropriate nonparametric tests (Wileoxin and Mann-Whitney) on the untransformed raw data. Each of the three t r e a t m e n t s DISC, SRC, and DRUG, had significant (P ~ 0.005) main effects in the predicted direction. With d/= 1/16, the Y ratios were 13.3 for D R U G , 114.7 for DISC, and 174.7 for SRC. All nine alcohol Ss had average error scores above the median for the placebo group. Within the alcohol and placebo groups, each S had higher scores for low DISC and for low SRC. When errors were distributed in a confusion matrix (see Conrad, 1964), we found t h a t for both the alcohol and placebo groups, a b o u t 800/0 of the errors were of the acoustic confusion variety. T h a t is, Ss in both groups tended to confuse consonants which contained similar vowel sounds (e.g., V, B, C). I n the pilot study mentioned earlier, the acoustic confusion errors were 85~ for the placebo and 83~ for the alcohol Ss. The interaction between the effects of DISC and SRC did not approach statistical significance in either the placebo or alcohol group. Thus, the effects of the 2 task variables were additive overall, confirming previous findings with R T as the dependent variable (Sternberg, 1969b ; Biederman and Kaplan, 1970). These effects were also additive within each level of the drug variable (i. e., the three-way interaction of DRUG, DISC, and SRC was not significant (F ~ 1.0). These results are consistent with the notion t h a t these two established treatments influence different stages of information processing in both sober and intoxicated Ss. The D R U G (alcohol-placebo) variable showed a significant (F1,16 = 19.5, P ~ 0.001) positive interaction with SRC, but the interaction between the effects of D R U G and DISC was not significant (F ~ 1.0). Fig.1 illustrates these results. Thus, moderately intoxicated Ss showed differential impairment under the condition of reduced S-R compatibility, but not under the condition of reduced stimulus discriminability.
V. K. Tharp Jr. et al.
38
6
,3
5
r~
High SRC o
2 i
1
0
i Placebo
I
Alcohol
I
I
Placebo
.....
Alcohol
Pig. 1. Effects of stimulus-response compatibility and stimulus-discriminability on errors of registration (square root transform). The graph on the left illustrates the main effect of stimulus response compatibility (SRC) and the interaction between the effect of that task variable and alcohol. The graph on the right shows that despite a powerful main effect of stimulus discriminability (DISC), the effects of alcohol and DISC were additive
Within the framework of the model, this finding suggests t h a t alcohol selectively impairs the stage of response selection-organization, but not the stimulus preprocessing stage. Experiment 2: Effects of Alcohol, Stimulus Discriminability, and Size of Memory Set on Reaction Time in a Memory Search Task To confirm the results of the first experiment and extend the analysis to a visual task, with R T as a dependent variable, we added a D R U G (alcohol-placebo) variable to the m e m o r y search paradigm described b y Sternberg (1967). The present experiment examined the combined effects of D R U G , DISC, size of m e m o r y set (MS) and response t y p e ("yes" or "no"). For the latter variable, the probalilities of " y e s " and " n o " were different. Sternberg (1966) found t h a t when Ss were required t o decide whether or not a probe stimulus was a m e m b e r of a memorized set of digits, for each unit increase in the size of MS, m e a n R T was increased by 38 msec. This average increase was independent of whether the appropriate response was " y e s " or "no". H e concluded t h a t the items in the memorized set were scanned (compared with the probe items) serially and exhaustively, with each item of MS requiring 38 msee to process. H e suggested t h a t the zero intercept of 397 msec for the regression of R T on size of MS represented the time needed to complete stages other t h a n those involved in scanning the MS items, namely stimulus preproeessing and response selection-organization.
Alcohol and Information Processing
39
As a p a r t i a l test for this l a t t e r n o t i o n , S t e r n b e r g (1967) varied 2 t r e a t m e n t s , d i s c r i m i n a b i l i t y of t h e probe s t i m u l u s (DISC) a n d size of MS. W i t h sufficient practice, low DISC a l t e r e d t h e i n t e r c e p t (q-63 msec), b u t n o t t h e slope, of t h e f u n c t i o n r e l a t i n g R T to MS. This a n d other evidence i m p l i e d t h a t degrading t h e probe s t i m u l u s i m p a i r s t h e s t i m u l u s preprocessing stage b u t n o t t h e serial m e m o r y processing stage. The results of our E x p e r i m e n t 1 lead to the p r e d i c t i o n t h a t t h e effects of DISC a n d D R U G will be additive. A significant i n t e r a c t i o n b e t w e e n t h e effects of D R U G a n d response t y p e w o u l d s u p p o r t t h e h y p o t h e s i s t h a t alcohol impairs t h e response selection-organization stage. A signifi c a n t i n t e r a c t i o n b e t w e e n D R U G a n d size of MS would implicate t h e s e q u e n t i a l m e m o r y processing (stimulus categorization) stage. Again, we did n o t a n t i c i p a t e a significant positive i n t e r a c t i o n b e t w e e n a n y of t h e established t a s k - r e l a t e d variables either i n sober or i n t o x i c a t e d Ss, n o r a t,hree-way or higher-order i n t e r a c t i o n i n v o l v i n g alcohol.
Method Subjects. The Ss were 24 male medical and graduate student volunteers from the University of Oklahoma Health Sciences Center, with normal vision and no history of drug abuse or heavy drinking, ranging in age from 21 to 38 years. They were paid for their participation. Upon arrival at the laboratory, each S was randomly assigned to one of two equal groups, alcohol or placebo, and was tested individually. During each of three sessions, S was penalized one point for each 0.i sec of I~T, l0 points for each error, and informed of his score for each 18 trial block. A $ 5.00 bonus was paid to the S with the lowest penalty score on each session. Apparatus. An Industrial Electronics Engineers, Inc. rear projection readout device produced white digits 1.5 cm high on a black 8 • 10 cm screen. On half of the trials a checkerboard grid (low DISC) was superimposed on the digits by simultaneously projecting horizontal and vertical white stripes (0.318 em wide) over the display. Pilot studies had shown that the low DISC condition increased RT by about 60 msec but without a notable increase in error rates. Stimulus presentation was controlled by a BRS tape reader system. Red and white lights located below the screen signaled incorrect and correct responses, respectively. White lights located at each edge of the screen were illuminated 1.25 see prior to stimulus presentation and served as a warning signal. The S was seated about 75 em from the screen, holding the index finger of each hand slightly above two microswitehes, with his palms resting on a table. Closure of the left microswitch always signaled a positive response (i. e., the digit on the screen was recognized as a member of MS). The period between onset of the probe digit and switch closure was recorded with an accuracy of • 1 msec by a General l~adio general-purpose counter (Type 1197). The experiment was conducted in a darkened room where tape recorded white noise, played at about 70 dB through loudspeakers, masked outside sounds. Procedure. The S came to the laboratory at the same time each day on 3 separate occasions spaced 3 days apart. The first and second sessions were practice and baseline runs, respectively. During the third session, prior to which S fasted for 4 hrs, he was given either placebo or ethanol. BAC measurements were taken 15 min after consumption of the final drink, during the middle of the 1-hr experimental session, and at the end. The mean BAC for the alcohol group was 98 mg-~ For individual Ss, the maximum variation in the 3 readings was about 25 rag-~ and on
40
V.K. Tharp Jr. et al.
average, between Ss the range was 85--120 mg-~ . All placebo BAC readings were well below 10 rag-~ . The fixed memory set paradigm used here is exactly like that described by Sternberg (1967). That is, a list of one, two, three, or four digits (taken from the ensemble 0 through 9) was memorized prior to presentation of a series of probe stimuli. The S's task was to decide whether or not each probe stimulus was a member of the MS and close the appropriate switch. Each trial consisted of the warning signal, display of a probe digit for 44 msec, S's response, and immediate feedback of accuracy. Trials were separated by an interval of 2.25 sec. Exactly the same procedures used by Sternberg (1967) to control for stimulus and response entropy, for systematic differences between the digits, for assigning all treatment combinations to each session, and for minimizing practice effects were used here and are only brieffly summarized below. Thus, the frequency of MS probe digits (4 in each block of 15 scored trials) and the frequency with which a particular digit occurred as a probe stimulus were constant throughout a session. Each session also contained every possible treatment combination of the four MS sizes and the two DISC conditions in a counterbalanced design. However, the treatment combinations involving th esubpart, MS = 3, were used for practice and thus always came first in each session. Each nonpractice combination of MS size and DISC consisted of 3 blocks of 18 trials each. For the analysis of RT, based on error-free data, practice effects were further minimized by eliminating the first block of 18 trials for any treatment condition and the first three trials of each of the two remaining blocks. Thus the RT analysis was based on scores from 30 trials for each combination of MS size and DISC. Within those 30 trials S was required to respond "no" 22 times and "yes" eight times. If the 15 scored trials in any 18 trial block contained more than 3 errors, the block was repeated. If the repeated block contained 3 errors or less, it was used for RT scoring; if it contained more than 3 errors, it was again repeated. Finally, only correct trials were used to compute mean RT. For the separate analysis of accuracy, error scores were computed from the second and third blocks of 18 trials within each nonpractice combination of DISC and MS size regardless of whether or not these blocks met the criteria for RT analysis.
Resutt8 Analysis o/Error Free Reaction Time Data. O n session three (DRUG), error rates i n t r i a l blocks accepted for R T analysis a v e r a g e d 0.9 per block i n t h e placebo group a n d 1.7 per block i n the alcohol group. However, i n order to o b t a i n r e l a t i v e l y error free blocks, it h a d b e e n necessary to a d d a t o t a l of 27 blocks of 18 trials for t h e 12 alcohol Ss a n d 8 blocks of 18 trials for t h e 12 placebo Ss. Analysis of v a r i a n c e for t h e error free R T d a t a from Session 3 rev e a l e d significant m a i n effects for DISC (F1,22 = 129.7, P < 0.001), MS (F1,22 ---- 110.0, P < 0.001), a n d response t y p e (F1,22 = 6.64, P < 0.05) ; b u t t h e t r e n d for D R U G was n o t significant a t t h e 0.05 level (F1,22---1.80, P > 0.10). The effects of DISC, MS, a n d response t y p e were addit i v e a n d n o n e of t h e t r e a t m e n t s i n t e r a c t e d with D R U G . Since the alcohol effect (between groups) on Session 3 showed o n l y a t r e n d t o w a r d statistical significance, difference scores were c o m p u t e d
Alcohol and Information Processing
E
550
Placebo
500
.-"
450 9~
41
Alcohol
...,
o.""/o/'Z
9
.....J"
400 350 -" 1"I
1
t
I
l
I
I
1
l
2
3
4
1
2
3
4
Memory Set Size
Fig.2. Regression of reaction time on size of memory set. Both graphs show the large overall difference in reaction tins (in milliseconds) between masked (dotted line) and unmasked (solid line) probe stimuli. Both graphs also show the large main effect of memory set size, and the additive relationship between the effects of stimulus diseriminability and memory set size. The dots along the regression lines are means. The regression equations are: Y = 324.4 4- 35.7 M and Y = 395.7 q- 30.8 M for unmasked and masked stimuli, respectively, for the placebo groups on the left. For the alcohol group on the right, the equations are: Y = 342.2 + 38.3 M for unmasked stimuli and Y = 397.9 -k 38.4 h[ for masked stimuli
between Sessions 2 a n d 3 for the Ss in each group. C o m p a r e d to the placebo group, average R T in the alcohol g r o u p increased 20 msec more from Session 2 (baseline) to Session 3 (DRUG). This difference was significant at the 0.05 level (t~2 = 2.27, 2-tailed). Thus, alcohol caused a small overall loss of speed. Fig.2 shows the regression of R T on MS size for each condition in Session 3. Linear regression a c c o u n t e d for more t h a n 99 ~ of the variance of m e a n R T s in every t r e a t m e n t combination, a n d the lateneies of positive a n d negative responses increased at a b o u t the same rate with increasing MS size. I n the placebo group with high discriminability (high DISC) of the probe digit, the slope of 37 msec and the zero-intercept of 321 msec correspond well t o Sternberg's (1967) estimates of 38 a n d 397 msec, respectively, for those parameters. F o r well-practiced Ss in Sternberg's study, degrading the probe stimulus (low DISC) increased the intercept b y 63 roses, a n d the slope b y less t h a n 3 msec. I n our well]practiced placebo Ss, the same t r e a t m e n t increased the zero-intercept b y a significant (tll = 6.75, P < 0.001) 61 msec. The slight reduction in slope associated with the m a s k was n o t statistically significant. As shown in Fig.2, the effects of MS a n d D I S C in the alcohol g r o u p were nearly identical to those in the placebo group. The v e r y small drug main effect was distributed a b o u t equally between intercept a n d slope.
42
V.K. Tharp Jr. et al.
These results confirm Sternberg's (1967) findings t h a t the effect of degrading the stimulus a n d increasing the size of the m e m o r y set are statistically independent, a n d t h a t neither t r e a t m e n t differentially affects t h e latencies for positive a n d negative responses. Moreover, none of these relationships were altered b y m o d e r a t e intoxication. Since alcohol h a d a v e r y weak effect on R T , the absence of interactions between t h a t t r e a t m e n t a n d other experimental variables m u s t be interpreted v e r y cautiously. Clearly the R T analysis on error free trials failed to disclose the site of the alcohol impairment. Analyses ol Error Data. Despite our efforts t o m a i n t a i n low error rates, intoxicated Ss (in trial blocks 2 a n d 3 of each condition) averaged a b o u t twice as m a n y errors as sober Ss. This effect consisted almost entirely of an increase in errors t o MS probe digits. ~ig.3 illustrates these results for the alcohol and placebo Ss. As can be seen, the intoxicated S t e n d e d to respond " n o " when he should have responded "yes". After converting error proportions for the alcohol group to the arcsin transform, analysis of variance within t h a t group showed significant m a i n effects for response t y p e (Fl,li ~ 108.7, P ~ 0.001), MS (Fi,ii 11.8, P ~ 0.01), a n d D R U G (Fi,li ~ 6.8, P ~ 0.05). There was a significant (Fi,ii ~ 9.2, P ~ 0.05) interaction between drug a n d response type. A n analysis of simple m a i n effects showed t h a t alcohol m a r k e d l y impaired a c c u r a c y only when the probe stimulus was a m e m b e r of the memorized list ( F l , i i ~ 17.24, P ~ 0.005). The effect of DISC on acc u r a c y was n o t significant, nor did t h a t t r e a t m e n t interact significantly with either MS or D R U G . None of the other interactions was statistically significant. Fig. 4. shows m e a n R T in the alcohol group 3 for correct and incorrect responses to MS probe digits, averaged across b o t h levels o f DISC. Performance was significantly faster for error trials t h a n for correct trials-the R T for incorrect responses to MS digits being 404 msee as c o m p a r e d with 465 msec for correct responses to MS digits (til -----4.46, P ~ 0.001) or 454 msec for correct responses to other digits (tli ~ 4.83, P ~ 0.001). 4 Since the pilot s t u d y summarized in f o o t n o t e 4 f o u n d no effect of left or right h a n d on R T , the m o s t parsimonious explanation for this 60 msec a d v a n t a g e of incorrect responses to MS probe digits would seem to be t h a t on those trials the alcohol Ss t r a d e d a c c u r a c y for speed. 8 A similar analysis comparing RTs for correct and incorrect trials was not performed for the placebo Ss because of their very low error rates. In an initial pilot study we examined the effect of handedness on "yes" responses. Five right-handed Ss were told to respond "yes" with their right hand during one session and with their left hand during a second session. Left-hand "yes" responses were a nonsignificant 5 msee faster than right-hand "yes" responses. Thus, nondrugged Ss showed no tendency to faster "yes" responses which could be ascribed to handedness.
Alcohol and Information Processing
43
.20 /
l"
l
.18
I
Alcohol / / /
.16
/ / / /
.14
/ / j,
o "t::
.12
/ / /
o
o t2.
.10
// o LJ
!
Placebo S
.08 .06 .04 .02 .00 " N o " Stimuli
"Yes" Stimuli
Fig. 3. Effect of alcohol on accuracy. This graph shows the main effect of response type {"yes" versus "no") on errors (proportion). Alcohol Ss made considerably more errors than placebos to stimuli for which they were required to respond "yes"
55O "8"
Correct Memory Set Responses 500
~
~ = 376.0 + 3&6M
400
~
350
"
e
rr
O' ~
|
" ct Memory Set Responses Y = 327.2 + 30.3M
~
t
"
~
,,~ ~ ' F
l
I
2
3
4
Size of Memory Set
Fig.4. Regression of reaction time (milliseconds) on size of the memory set in alcohol Ss. The regression functions illustrate that incorrect responses to memory set probe stimuli were about 60 msec faster than correct responses. :For each of the 12 Ss, the intercept for incorrect responses was lower than for correct responses and the overall difference in intercepts of 49-msec was significant beyond the 0.01 level. Dots along the regression lines represent means
44
V.K. Tharp Jr. et al.
On error trials, the intoxicated S was probably not simply responding to the onset of the probe stimulus, disregarding its content. The 404 msec required for incorrect responses to positive probe digits is approximately double that usually reported for simple visual R T (about 200 reset). Thus, it appears that the intoxicated S did partially process the probe digits on incorrect trials. However, the 50--60 msec advantage suggests that one of the stages in the information processing sequence may have been omitted on those trials. What stage could have been skipped or impaired when the intoxicated S responded " n o " to MS probe digits? If, for these error trials, we consider the regression of R T on MS size, a reduced slope, but unaltered intercept, would imply that all or part of the serial memory processing stage was omitted. Conversely, a reduced intercept but unaltered slope would imply t h a t some other stage, ostensibly reflected in the intercept (i.e., stimulus preprocessing or response selection-organization) had been omitted. Fig.4, based on data from the DRUG session for the alcohol Ss, shows that the intercept was reduced considerably more than the slope on incorrect responses to positive MS probes. On those trials the intercept was 49 msec lower (tll ---- 2.46, P ~ 0.01) than for correct responses, but the slope was not significantly reduced. Thus, the impaired or missing stage is probably reflected in the intercept. Within the framework of the additive model, stimulus preprocessing could be tentatively ruled out as the focus of alcohol effects if R T on masked trials increased systematically with stimulus degradation. The mask caused a significant increase in total RT on both correct and incorrect responses to MS probes. Moreover, the differential effects of DISC on correct and incorrect trials were not statistically significant (t < 1). Thus, it is likely that stimulus preprocessing was essentially intact on these incorrect trials, B y elimination, these results implicate the stage of response selection-organization as the most likely target for alcohol effects. On about 200/0 of trials in which the probe digit was a member of the MS, the intoxicated S apparently preprocessed and encoded the probe digit and compared it sequentially with the MS items. However, prior to execution of the response, he omitted the essential set of cognitive operations required for selection and organization of the response. W h y did this happen? Since the experiment contained a systematic bias toward the response " n o " (p ~ 11/15), the intermittent failure of the response-selection stage in the intoxicated S may have been due in part to the surprisal value of the low probability "yes" response. One purpose of Experiment 3 was to examine the RT and error performance of intoxicated Ss in blocks of trials within which response probabilities were equal.
Alcohol and Information Processing
45
Experiment 3: Effects of Alcohol, Stimulus Diseriminability, Number of S-R Alternatives and S-R Compatibility on Verbal Reaction Time The first 2 experiments implicated as one locus of alcohol impairment, t:hose operations in the serial additive model associated with selection and organization of the response. However, conclusions derived from Experiments 1 and 2 depended on the analysis of error data with rather low base rates. Since Moskowitz and R o t h (1971) and t I u n t l e y (1972) had found systematic effects of alcohol on verbal R T in somewhat more traditional 1 : 1 mapping tasks, we shifted to similar procedures for Experiment 3. Sternberg (1969b, Experiment 5) reported a suitable paradigm in which the effects of DISC, SI~C and stimulus response uncertainty (SI%U, number of stimulus-response alternatives) on verbal R T were examined. H e reported t h a t DISC and SRC had additive effects on RT. The effects of SRU showed a weak positive interaction with DISC and a strong interaction with SI%C. H e concluded t h a t although SRU probably had some influence on the stimulus preprocessing stage, it had its greatest effect upon response selection-organization. The results of our Experiments 1 and. 2 permitted a firm prediction t h a t D R U G would not interact with DISC. Instead, if the principal targets of alcohol are cognitive operations involved in response selectionorganization, the effects of the drug on verbal R T should show positive interactions with SRC and SRU. Furthermore, a significant three-way interaction between DRUG, SRC and S R U was anticipated. These predictions were strengthened b y I t u n t l e y ' s (1972) finding of a significant alcohol b y SRU interaction. Additivity between D R U G and DISC in the presence of Sternberg's " w e a k " interaction between SRC and DISC would strongly support the view t h a t alcohol specifically impairs the response selection side of the i~aformation processing system. Finally, if the MS errors found with alcohol in Experiment 2 were due to the relatively low probability of MS probe digits, then there should be no such error trend within the equalprobability sets of Experiment 3; but alcohol should produce greater impairment with the increased number of alternatives.
Method Subjects. The Ss were 18 male students from the University of Oklahoma Health Sciences Center and Oklahoma City University, with normal vision and no history of heavy alcohol or drug abuse, who ranged in age from 21 to 35 years. They were paid for participation in four experimental sessions and could earn additional bonus pay for exceptional performance. The payoff and penalty procedures were identical to those of Experiment 2, except that in the attempt to further reduce error rates, the penalty for each error was raised from i0 to 20 points. Ss were tested individually.
46
V . K . Tharp Jr. et al.
Apparatus. Experiment 3 used the same display system as Experiment 2, but the RT apparatus was modified to measure verbal RT. S's response, detected by a microphone, was amplified, clipped, integrated, and let to a level detector which acted to stop the electronic counter. There was constant 14 msec delay from onset of sound to triggering of the level detector. Differences in speech loudness and phonetic characteristics of the response had negligible effects on triggering latency (about 2 msec). Feedback for incorrect responses was accomplished by a red light located beneath the display screen, and was delivered manually. As in Experiment 2, S was seated in a darkened room, and a low level of white noise masked outside sounds. Procedure. Singe digits were presented visually and sequentially in blocks of 20 with an intertrial interval of 2.25 sec. The S responded either by repeating the digit presented (high SRC) or the next digit in ordinal succession (low SRC). In a given block of 20 trials, the digits were randomly selected from either two (low SRU) or eight (high SRU) alternatives, with the constraint that all alternative digits were equally represented in the last 16 trials. The eight alternatives consisted of all single digits except 0 or 9, while the two alternatives were selected randomly from the set of eight. The stimuli were either masked (low DISC) or unmasked (high DISC) using the procedure of Experiment 2. The two levels of each of the three treatments were varied systematically in each session. All treatment combinations involving high DISC were presented prior ~o any combination involving low DISC. Similarly, both levels of SRC involving two alternatives were presented before any conditions with eight alternatives. The first four trials of each 20-trial block were used for warm-up and discarded from the analysis. In addition, with five blocks for each treatment combination, only the last four were used for data analysis. This was a within-Ss experimental design in which each S participated in four experimental sessions lasting about 2 hrs each, including a 5-min break midway through the session. For each S, the experimental sessions were scheduled at the same time each day, with 1 day intervening between sessions 1, 2 and 3, and 2 days between sessions 3 and 4. Sessions 1, 2, and 4 provided baseline data, whereas on Session 3, each S consumed 1.32 ml of 95~ ethanol per kg body weight. As in the previous experiments, the ethanol was mixed 1:4 with ginger ale or orange drink, divided into three drinks and consumed in 30 min. Breathalyzer readings were taken 15 min after consumption of the drinks, in the middle of the task, and at the end of the session. The means for the 3 BAC readings were not significantly different from one another and averaged 96 mg-~ (range 85--110 rag-C/0). Results T h e r e were practice t r e n d s t h r o u g h all four e x p e r i m e n t a l sessions. Thus, R T s were significantly faster o n D a y 2 t h a n D a y 1 (t17----3.76 P ~ 0.01). However, the overall i m p r o v e m e n t from D a y 2 to D a y 4 was o n l y 11.8 msee (tl~ ~ 1.84, P ~ 0.10). Therefore, the d a t a from the second a n d f o u r t h sessions were averaged for a baseline score. As a n t i c i p a t e d , with e q u a l l y p r o b a b l y s t i m u l i a n d increased penalties, error frequencies were v e r y low. E v e n t h o u g h u n i n t e l l i g i b l e responses a n d r e s p o n s e s b e g i n n i n g with the wrong s o u n d (e.g., " u h - - t h r e e " ) were considered incorrect, overall error scores on the alcohol session averaged o n l y 1.1 ~ or a b o u t one error per 5 blocks o f trials. As a result, the R T analysis, based o n error free d a t a , c o n t a i n e d n e a r l y all of the trials. Since
Alcohol and Information Processing
47
6O E. t--
50 40
~' 3(? ._c ~ 2O =
10
~
0
2
8
SR Uncertainty
High Low SR Compatibility
Fig. 5. Increase in reaction time from baseline to alcohol as a function of stimulusresponse uncertainty and stimulus-response compatibility. Bar graphs are presented in order to eliminate the large main effects of stimulus-response (SR) uncertainty and SR compatibility which tend to obscure the interactions. Baseline is represented by a zero increase in reaction time (RT). Graph on the left illustrates that alcohol increased the average RT associated with 8 alternatives by 13 msec more than that with 2 alternatives (P < 0.01). Graph on the right illustrates that alcohol hlereased the RT with low SR compatibility by 14 msec more than with high SR compatibility (P < 0.Ol)
t h e m e a n error r a t e p e r t r e a t m e n t c o m b i n a t i o n was less t h a n one p e r S, no s y s t e m a t i c analysis o f errors or t h e i r a s s o c i a t e d l~Ts was c o n d u c t e d . E a c h o f t h e four t r e a t m e n t s h a d significant m a i n effects. Alcohol c a u s e d a n a v e r a g e increase in v e r b a l R T o f 55 msee (F1,17 = 37.2, P < 0.001) over t h e baseline average. T h e high SI~U c o n d i t i o n caused, on t h e a v e r a g e , a 54 msec increase o v e r low S R U (F1,17 = 167.1, P < 0.001); low SRC i n c r e a s e d m e a n R T 94 msec (F1,17 ~ 266.3, P < 0.001), a n d low D I S C i n c r e a s e d m e a n R T 4 8 m s e c (F1,17 = 119.3, P < 0.001). As e x p e c t e d , t h e effects o f SRC a n d S R U showed a significant p o s i t i v e i n t e r a c t i o n (F1,17 = 744.4, P < 0.001), s u p p o r t i n g S t e r n b e r g ' s (1969b) conclusion t h a t t h e s e t w o s t a g e s b o t h influence r e s p o n s e selectiono r g a n i z a t i o n . T h e r e was no t r e n d in t h e s e d a t a , however, for t h e w e a k b u t significant i n t e r a c t i o n r e p o r t e d b y S t e r n b e r g b e t w e e n t h e effects o f D I S C a n d S R U (Fl,t7 < 1.0). F i n a l l y , t h e r e was a significant negative i n t e r a c t i o n b e t w e e n D I S C a n d SRC (FI,t7 = 12.4, P < 0.005), a n une x p e c t e d r e s u l t for which we h a v e no i n t e r p r e t a t i o n . As i l l u s t r a t e d in Fig. 5, t h e effects of D R U G s h o w e d a significant p o s i t i v e i n t e r a c t i o n w i t h b o t h S R U (P1,]7 = 12.9, P < 0.005) a n d S R C
48
V.K. Tharp Jr.
et al.
($'1,17 ~ 11.7, P ~ 0.005). However, the anticipated three-way interaction between DRUG, SRU and SRC was just short of significance at the 0.05 level (F1,17 ---- 3.9, P ~ 0.10). As in Experiments 1 and 2, the effects of DRUG and DISC were additive (F1,17 ~-- 1.03, P ~ 0.25). No other interactions approached significance (F being less than 1.0 in each case). Thus, the results of this experiment are consistent with those of Experiments 1 and 2. The effects of alcohol interacted with those of treatments that probably influence the stage of response selection-organization, but not with the effects of DISC, the treatment which probably influences the stimulus preprocessing stage. Discussion The 3 experiments reported here, employing auditory or visual stimuli in character-recognition tasks, used both error rates and reaction time to measure performance in sober and moderately intoxicated Ss. Stimulus diseriminability and at least one other established task-related treatment were varied in each experiment in an a t t e m p t to influence, specifically, three stages of information processing: stimulus preprocessing, stimulus categorization (memory scanning), and response selectionorganization. Despite the use of a variety of independent and dependent variables, the results were remarkably consistent within both alcohol (average B A C ~ 100rag-~ ) and placebo-control conditions. More specifically patterns of additivity and interaction between the task-related experimental treatments were similar for both error data and RT, and were identical in sober and intoxicated states. These results imply that the serial, additive-stage information processing model from which the experiments and associated treatments were derived is a useful paradigm for either speed or accuracy scores obtained either in normal or altered psychophysiologieal states. Experiment 1, which required written transcription of acoustically presented letters, used the task-related established treatments stimulus discriminability (DISC) and stimulus-response compatibility (SRC) with accuracy as the dependent variable. DISC varied the signal-to-noise ratio of the tape-recorded letters, whereas SRC altered the difficulty of mapping the stimulus on the response. Each treatment increased average error rates, but as in normal Ss with R T as the dependent variable, their effects were additive in both the sober and intoxicated state. Thus in both states, DISC and SRC appear to influence different stages of information processing, probably stimulus preprocessing on the one hand, and response selection-organization on the other (Sternberg, 1969 b; Biederman and Kaplan, 1970). The memory-search R T task in Experiment 2, taken from Sternberg (1967), used visually presented probe digits and a binary manual re-
Alcohol and Information Processing
49
sponse. The treatments DISC (masking of the probe digit), size of MS (fixed memory set), and response type ("yes" or "no") were varied with both RT and accuracy as dependent variables. As anticipated, the three task-related variables had additive effects on speed in both sober and intoxicated Ss, and on accuracy in the alcohol group. Because of very ][ow error rates, accuracy effects were not examined in the placebo group. These results support Sternberg's (1967) contention that the three treatments influence different, non-overlapping stages of information processing; probably stimulus preprocessing, stimulus categorization and response selection-organization, respectively. Experiment 3, a i : 1 mapping task which required verbal responses to visually presented digits, varied DISC (as in Experiment 3), SRC (as in Experiment 1) and SRU (number of equally likely S-R alternatives) with verbal R T as the dependent variable. The strong positive interaction found between SRC and SRU in both sober and intoxicated states implies that these task-related variables influence the same stage of information processing; i.e., response selection-organization. As Sternberg (1969b) reported, the effects of DISC and St~C showed no positive interaction. The small but statistically significant negative interaction found here is probably not interpretable within the additive model of information processing. Sternberg's (1969 b) weak positive interaction between DISC and SRU was not confirmed. Taken together, the results of these experiments recommend for both sober and intoxicated Ss a serial additive model for processing of simple stimuli consisting of at least three non-overlapping stages: stimulus preproeessing, stimulus categorization and response selection-organization. Which of these stages is most vulnerable to alcohol ? In Experiment 1, with accuracy of transcription as the dependent variable, alcohol increased the frequency of acoustic confusion errors, i.e., such substitutions as "V" for "B". With this result, suggesting impairment on the perceptual side, one might have anticipated a positive interaction between the effects of Dt~UG and DISC. However, the drug variable showed a strong positive interaction with S-I~ compatibility and no trend toward an interaction with DISC. That is, the intoxicated S showed differential impairment in a treatment condition which increased the difficulty of responding, not with one which degraded the stimulus. This finding suggested that ~dthin the framework of the model, at least one locus of alcohol effects is the stage of response selcetion-organitazion. In the memory-probe task of Experiment 2, a test of this hypothesis was weakened by the fact that alcohol caused very little impairment of speed, the principal dependent variable. For t h a t performance measure the drug variable showed no positive interaction with any of the established task-related treatment variables. Despite the use of penalty points 4 Psychopharmacologia (Berl.), Vol. 40
50
V.K. Tharp Jr. et al.
and financial rewards, alcohol Ss made a b o u t twice as m a n y errors as controls. This effect, consisting almost entirely of an increase in " n o " responses to positive probe digits, resulted in a strong positive interaction between D R U G and response type. Thus, with accuracy as the dependent variable, alcohol impairment again appeared on the response side. Further analysis revealed t h a t the RTs for incorrect responses were 50--60 msec faster t h a n for correct responses, suggesting t h a t on those trials a stage of information processing might have been omitted. Examination of the slope and intercept parameters of the regression function relating R T to size of the m e m o r y set indicated t h a t the 50--60 msee advantage on incorrect trials was accounted for primarily b y a reduction in the intercept. Theoretical considerations implicate the slope of the regression function in the serial m e m o r y scanning process, and the intercept in at least two other stages, stimulus preprocessing and response selection-organization. Because the mask (low DISC) showed its expected effects on error-related RTs, we concluded t h a t the omitted stage was probably response selection. Thus, the data were consistent with the notion t h a t on the incorrect trials, the inebriated S both preprocessed and categorized the stimulus (as a member of MS). However, prior to executing the response, S omitted the essential set of cognitive operations required for selection of the correct response (the stage of response selection-organization) quite as if he had made a slip of the tongue. We suspected t h a t the biased error distribution found with alcohol in Experiment 2 resulted from a systematic bias toward the response " n o " which was p a r t of the experimental design. I n E x p e r i m e n t 3, with unbiased response probabilities and increased penalties for errors, average error rates in both the sober and intoxicated state were very close to zero. I n this study, alcohol caused a systematic increase in verbal R T and the effect of D R U G showed strong positive interactions with two associated established treatments, S-R uncertainty and S-R compatibility. As in Experiment 1 and 2, the effects of D R U G and DISC were additive. Thus, again the intoxicated S showed differential deficit under t r e a t m e n t conditions which increased the difficulty of responding, but not with one t h a t impaired perception of the stimulus. This latter result confirms t t u n t l e y ' s finding of a positive interaction between the effects of alcohol and SRU. I t will be recalled t h a t SRU and SRC themselves show strong positive interactions both in Sternberg's (1969b) studies and ours, and t h a t theoretical consideration implicate both t r e a t m e n t s in the stage of response selection and organization. I n summary, these three experiments strongly support the notion t h a t in character-recognition tasks, one i m p o r t a n t source of impairment with moderate alcoholic intoxication is a difficulty in selecting and organizing the correct response. This deficit m a y be manifested either b y an increase
Alcohol and Information Processing
51
in r e a c t i o n t i m e or in errors or b o t h , d e p e n d i n g on such c o n d i t i o n s o f t h e e x p e r i m e n t as differential response p r o b a b i l i t i e s a n d t h e payoffs for s p e e d a n d a c c u r a c y . T h e r e was no e v i d e n c e in a n y o f t h e e x p e r i m e n t s t h a t p e r c e p t u a l o p e r a t i o n s such as s t i m u l u s p r e p r o c e s s i n g or e n c o d i n g were i m p a i r e d b y BACs of a p p r o x i m a t e l y 100 mg-~ . Since t h e alcohol t r e a t m e n t c o n d i t i o n d i d n o t a l t e r t h e p a t t e r n s of a d d i t i v i t y a n d i n t e r a c t i o n a m o n g t h e t a s k - r e l a t e d t r e a t m e n t variables, i t is a s s u m e d t h a t t h e d r u g d i d n o t cause q u a l i t a t i v e a l t e r a t i o n s in t h e w a y i n f o r m a t i o n was processed. This implies t h a t t h e serial a d d i t i v e m o d e l which p r o v i d e d t h e c o n t e x t for t h e e x p e r i m e n t s is r o b u s t to d r u g - a l t e r e d s t a t e s of consciousness, a n d t h a t i t m a y be a p p l i c a b l e t o t h e s t u d y o f i m p a i r e d p e r f o r m a n c e a s s o c i a t e d w i t h o t h e r a l t e r e d s t a t e s such as t h o s e i n d u c e d b y sleep dep r i v a t i o n , b r a i n d a m a g e or p s y c h o p a t h o l o g y . This study was supported in part by the Office of the Surgeon General, Department of the Army, Research Contract No. DADA-17-73-C-3157 and by USPttS l~esearch Grant No. AA00212-06. Part of the data were reported at the 20th International Institute on the Prevention and Treatment of Alcoholism in Manchester, England. We thank Lawrence C. Cowden, Sharolyn Lentz, Judy Smith, Joe Gold, Jan Staat and Cindy Coulter for their technical assistance, management of subjects and analysis of data.
References Biederman, I., Kaplan, R.: Stimulus discriminability and S-~ compatibility: evidence for independent effects in choice reaction time. J. exp. Psychol. 86, 434--439 (1970) Darley, C. F., Tinklenberg, J. R., Hollister, T. E., Atkinson, R. C. : Marihuana and retrieval from short-term memory. Psychol?harmacologia (Berl.) 29, 231--238 (1973) I-Iuntley, M. S. : Influences of alcohol and S-R uncertainty upon spatial localization time. Psychopharmaeologia (Berl.) 27, 131--140 (1972) Jones, B. M. : Memory impairment on the ascending and descending limbs of the blood alcohol curve. J. abncrm. Psychol. 82, 24--32 (1973) Moskowitz, H., Burns, M. : Effect of alcohol on the psychological refractory period. Quart. J. Stud. Alcohol 82, 782--790 (1971) Moskowitz, H., DePry, D. : Differential effect of alcohol on auditory vigilance and divided attention tasks. Quart. J. Stud. Alcohol 29, 54--63 (1968) Moskowitz, H., Roth, S. : Effect of alcohol on response latency in object naming. Quart. J. Stud. Alcohol 3~o, 969--975 (1971) :Norman, D.A. (ed.): Models of human memory. New York: Academic Press 1970 Rutschmann, J., Rubenstein, L. : Time estimation, knowledge of results and drug effects. J. psychiat. Res. 4, 107--114 (1966) Smith, E. E. : Choice reaction time: an analysis of the major theoretical positions. Psychol. Bull. 69, 77--110 (1968) 4*
52
V . K . Tharp Jr. et al.
Sternberg, S. : High speed scanning in human memory. Science 158, 652--654 (1966) Sternberg, S.: Two operations in character recognition: Some evidence from reaction time measurements. Perception and Psychophysics 2, 45--53 (1967) Sternberg, S. : Memory scanning: mental processes revealed by reaction time experiments. Amer. Sci. 57, 421--457 (1969a) Sternberg, S. : The discovery of processing stages. Extensions of Donder's method. Acta psychol. (Amst.) 80, 276--315 (1969b) Van K. Tharp Jr. Department of Psychiatry University of Oklahoma Health Sciences Center P. O. Box 26901 Oklahoma City, Oklahoma 73190, U.S.A.