Crossmodal divided attention in rats: effects of chlordiazepoxide and scopolamine

Psychopharmacology (1994) 115:213-220 Psychopharmacology © Springer-Verlag 1994

Crossmodal divided attention in rats: effects of chlordiazepoxide and scopolamine Jill McGaughy, Janita Turchi, Martin Sarter Department of Psychology, The Ohio State University, 27 Townshend Hall, Columbus, OH 43210, USA

Received: 12 April 1993 / Final version: 29 October 1993

Abstract. "Divided attention" is a psychological con- struct that hinges on assumptions about a fixed finite capacity of subjects to simultaneously process multiple sets of information. A model of a crossmodal divided attention task was developed in rats. Initially, rats were trained consecutively in operant auditory and visual conditional discrimination tasks. The final task consisted of two successive blocks of 20 trials per modality (modality certainty), followed by 60 trials comprising a semi-randomized sequence of stimuli of both modalities (auditory or visual) and qualities (flashing/pulsing or constantly turned on; modality uncertainty). In comparison to unimodal blocks of trials, performance in the mixed condition was assumed to reflect the demands on the parallel processing of two sets of stimulus-response rules. While response accuracy remained unchanged, response latencies were generally longer in the bimodal condition. Ad- ministration of scopolamine (0.03, 0.06, 0.1 mg/kg) or chlordiazepoxide (1, 3, 5, 8 mg/kg) dose-dependently increased response latencies. The scopolamine-induced increase in response latencies was greater in the mixed condition. Cost-benefit analyses demonstrated that the absolute divided attention costs (in ms) were generally higher for visual than for auditory stimuli. Both drugs produced qualitatively similar effects; however, scopolamine was more potent in increasing the absolute divided attention costs than chlordiazepoxide. These data are discussed in terms of the validity of this animal paradigm, and of hypotheses about the effects of benzodiazepine receptor agonists and muscarinic antagonists on brain information processing capacity.

Key words: Divided attention - Scopolamine - Chlor- diazepoxide - Rat

Although "divided attention" is widely understood as the ability of the subject to "do two things at the same time"

Correspondence to: M. Sarter

(Pashler 1993), it has also been speculated that what is perceived as parallel processing may rather be a result of continuous shifts between multiple activities ("time sharing"). The intense theoretical debate about this con- ceptual dichotomy appears to exaggerate the significance of determining the exact cognitive mechanisms that mediate dual task performance (Hirst 1986; Allport 1989). Both views (i.e., parallel versus serial) hinge on the intu- itively attractive hypothesis that the capacity of brain information processing systems is limited; thus, similar performance decrements may result from either an ex- haustion of the capacity for simultaneous processing or the allocation of processing resources.

As already observed by William James (quoted in Hirst 1986), the extent to which limited processing resources are taxed by a certain task depends largely on stimulus characteristics (e.g., modality), internal codes (e.g., visual versus verbal), response characteristics (e.g., habitual motor responses or skills versus newly trained responses or speech), and levels of processing (reflexive or automatic versus controlled; Schneider and Shiffrin 1977; see also Wickens 1984). These considerations allow one to predict that, if dual task performance is based on well-trained responses (skills) or largely automated processing of information (as typically results in animals that require extensive training to gain stable performance; Sarter 1990), the effects of the division of attention will be primarily revealed by the increase in the time taken to shift processing resources between the different tasks, rather than by performance accuracy. In relevant experiments in humans, no effects on response accuracy were found in "divided" conditions (McDowd and Craik 1988; khlkenstein et al. 1991; Hohnsbein et al. 199t); rather, "latency becomes the relatively more reliable index of decision-making quality.....since errors are fewer, they provide a less reliable source of information concerning quality" (Wickens 1984, p. 336).

Divided attention has been hypothesized to represent a crucial element in the cognitive impairments associated with various neuropathological processes, and a major intervening variable in the cognitive effects of drugs. The

214

ava i lab le evidence suggests tha t i m p a i r m e n t s in d iv ided a t t en t ion represents the mos t r obus t cogni t ive change assoc ia ted with n o r m a l ag ing (e.g., Wr igh t 1981; Cra ik and Byrd 1982; M c D o w d and Cra ik 1988). Older ind iv id- uals a p p e a r to spend much of their p rocess ing capac i ty for the o rgan i za t i on of the t ime shar ing process , hav ing re la t ively litt le res idual capac i ty for s t imulus - re la ted p ro - cessing. Similarly, pa t i en t s with Alzhe imer ' s disease were consis tent ly found to show grea te r i m p a i r m e n t s in divided a t t en t ion tasks when c o m p a r e d with focused or directed a t t en t ion tasks (e.g., G r a d y et al. 1989; F i lo teo et al. 1992). Fur the rmore , the amnes ic effects of benzodi - azepine recep to r (BZR) agonis ts have been a t t r i bu ted to d rug - induced reduc t ions in the capac i ty of the "centra l executive" (Cur ran 1991a; Badde ley 1992). Similar ly, the cogni t ive effects of the muscar in ic r ecep to r b locke r scopo lamine , which a p p e a r qua l i ta t ive ly c o m p a r a b l e with the effects of B Z R agonis ts (Pres ton et al. 1988; Rus ted et al. 1991), have been a t t r i bu t ed to d rug - induced impa i r - ments in the a l loca t ion of a t t en t iona l resources (e.g., D u n n e and H a r t l e y 1986; Wesnes et al. 1988; C u r r a n et al. 1991). To invest igate the neu rona l mechan i sms medi - a t ing in tac t and i m p a i r e d d iv ided a t ten t ion , a n d the potent ia l of p h a r m a c o l o g i c a l s t ra tegies for the t r ea tmen t of such impa i rmen t s , an ima l behav io ra l tests of d iv ided at - t en t ion are required.

In con t r a s t to tests for sus ta ined a t t en t ion (Dud- chenko et al. 1992; M o o r e et al. 1992a), the bas ic charac - terist ics of the dua l t a sk p rocedure s used in h u m a n s to test d iv ided a t t en t ion are no t read i ly m o d e l l e d in rodents . The s imul t aneous t e m p o r a l p rocess ing p r o c e d u r e used by Ol ton et al. (1988) to inves t iga te f ronta l cor tex func- t ions represents an exception. However , the measures ob- ta ined f rom this p a r a d i g m (i.e., the prec is ion of t iming of two stimuli) a p p e a r r a the r difficult to in te rpre t in te rms of the effects of the d ivis ion in a t t en t ion on more conven- t ional measures of pe r fo rmance (latency, accuracy). Therefore, in this exper iment , an ope ran t t a sk was de- s igned using the key c o m p o n e n t s of a c ro s smoda l d iv ided a t t en t ion task used for test ing h u m a n s (Fa lkens te in et al. 1991; H o h n s b e i n et al. 1991). This t a sk requi red the hu- m a n subjects to e i ther detect and identify two let ters presented in one m o d a l i t y (focused a t t en t ion ; m o d a l i t y certainty), o r with equa l p r o b a b i l i t y as visual or a u d i t o r y s t imuli (d ivided a t t en t ion ; m o d a l i t y uncer ta inty) . The longer response latencies obse rved in the d iv ided a t ten- t ion cond i t i on were a t t r i bu t ed to the requi red d iv is ion in a t t en t ion be tween the two modal i t ies . Add i t iona l ly , in the d iv ided a t t en t ion cond i t ion , evidence for an asym- met r ica l a l loca t ion of resources in favor of the visual m o d a l i t y was found. In the presen t exper iment , an ima l s were t r a ined no t on ly to d i v i d e a t t en t ion with respect to the de tec t ion and ident i f ica t ion of s t imul i of two m o d a l i - ties, bu t add i t i ona l ly to d iver t thei r p rocess ing capac i ty be tween discrete s t imulus - response rules for a u d i t o r y and visual stimuli.

Materials and methods

Subjects. The study commenced with 12 male Long-Evans Hooded rats (Harlan-Sprague-Dawley, Inc., Indianapolis, Ind.) weighing

200-224 g. The animals were individually housed in a humidity and light controlled room, with lights on at 6:00 a.m. for a 12L:12D pattern. The animals were extensiveIy handled for 2 weeks prior to testing. Animals were moderately food deprived (10% weight loss) while water was available ad libitum.

Apparatus. Twelve operant chambers (MedAssociates, East Fair- field, Vt.) equipped with two retractable levers, a houselight (2.8 W), a food pellet dispenser, a 2900 Hz sonalert tone generator, and three panel lights (2.8 W) were used. The panel lights and levers were located on the wall opposite to the pellet dispenser, requiring the animal to turn around after receiving reinforcement in order to face the intelligence panel. Records of stimulus presentation, lever oper- ation, and food pellet delivery were computerized using a 386 PC clone and the Med-PC software (MedAssociates, East Fairfield, Vt.).

Behavioral training. Training occurred between 2:00 p.m. and 6:00 p.m., 6 days per week. Rats were first trained to bar press using an FR1 schedule (Noyes Pellets, 45 mg). Once the animals had received at least 50 reinforcements for 2 consecutive days, they were trained in a visual or auditory conditional discrimination task (six animals started with the visual task, six with the auditory). In either modality, stimuli were presented in a pseudo-randomized order and were either constantly turned on ("constant") or flashing/pulsing (on-off at 20 Hz) tones or lights. Visual stimuli were presented at the central panel light. For the visual conditional discrimination task, correct responses were those which conformed to the following rule: light constant = left lever, light flashing = right lever. Correct responding in the auditory shaping task followed: tone constant = right lever, tone pulsed = left lever. For each of the above rules, half of the animals were trained on the converse rules, tn pilot experiments, all combinations of rules were tested; however, animals trained in accordance with pairs of rules that included the same correct lever for the same stimulus quality (e.g., the right lever for flashing light and for pulsing tone) were found to learn the task rapidly, and no effects of the random presentation of stimuli of the two modalities were found. It was assumed that this finding was a result of a generalization to stimulus quality; thus, there are only two pairs of rules that prohibit generalization to stimulus quality, and they are based on opposite correct levers for stimuli of different modality but comparable quality. Each training session commenced with illumi- nation of the houselight and extension of both levers. Levers were activated at the onset of stimulus (presented for 10 s) and remained active throughout stimulus presentation, Correct responses caused the delivery of a pellet which was followed by a variable intertrial interval (12 _+ 4 s). An incorrect response, or a failure to respond to the stimulus was succeeded by the ITI. Lever operations during the ITI and in the absence of stimuli had no scheduled consequences. The initial criterion for switching between stimuli of one modality to the next was a response accuracy of above chance level, and no fewer than 50 total responses.

Animals were trained in these unimodaI conditional discrimina- tions until they were attaining the criterion for switching modatities on a daily basis, i.e., within one session. Once they were performing at this level, the animals were moved to a second shaping task. This task was composed of four successive blocks of 25 trials each of stimuli of one modatity. Criteria were set such that when an animal was performing consistently at the level of > 70% correct responses over the 100 total trials, and evenly on both levers across all sets, it progressed to the final task. The animals were transferred to the training in the final task following 69 (mean) sessions of training in these preliminary versions of the task.

The shaping tasks prepared the animals for a final task involv- ing the following rules for selection of the correct response lever: light constant = left lever, light pulsed = right lever, tone constant = right lever, and tone pulsed = left lever. Training for the above rules as well as the converse set was balanced across the group. The final task consisted of a total of 100 trials and was composed of two successive unimodal blocks of 20 trials each (subsequently termed "unimodal condition"), followed by 60 trials of pseudo-random presentation of all stimuli (subsequently termed

215

1. Blocks of unimodal trials:

I V_ _.,

2. Semi random order of: ZIO or ~ 0 or l l O o r --10

Fig. 1. Schematic illustration of the behavioral task. The central panel light (circle) or the sonalert tone generator (box on the upper left site of the box) were either turned off (empty), constantly turned on (black) or flashing (on-off at 20 Hz; pattern). A daily session started with blocks of unimodal trials (i.e., 20 trials of visual stimuli followed by 20 trials of auditory stimuli, or vice versa). These blocks were followed by 60 trials consisting of a semi-randomized sequence of flashing or constant tones or lights. Arrows indicate the correct lever in response to each stimulus type (half of the animals were trained in accordance with the alternative combination of correct responses). Note that, for example, different levers were correct in response to flashing light and pulsing tone, i.e., correct responding could not be generalized to stimulus quality

"bimodal condition". Initial experiments demonstrated that with- out these two unimodal ("priming") blocks preceding the bimodal condition, performance in the bimodal condition disintegrated below chance level. Thus, the sequence of unimodal blocks followed by bimodal trials was kept consistent over all experiments. While the available data (see Results) do not suggest that performance declined in the course of a daily session, it cannot be excluded that the effects of the drugs were confounded by the effects of the sequence of uni- and bimodal blocks of trials (see Discussion). The order of the two unimodal priming blocks was initially switched on a daily basis and then on a double alternation basis, as this allowed for greater stability of performance (see Fig. 1 for an illustration of the elements of the final task).

Once the animals reached a stable baseline of approximately 70% correct responses for each stimulus type across the unimodal blocks (no less than 57% correct responses for any one stimulus type), drug testing was started. The average number of training sessions from commencement of final task training to drug testing was 23 (i.e., the average number of sessions/days was 92 from the beginning of the experiment to drug testing). Six out of the original 12 animals reached the final stage of testing (i.e., met all criteria) and their data were included in the data analyses.

Drugs and drug administration procedures. Drugs tested included scopolamine hydrobromide (Sigma; 0.03, 0.06, 0.1 mg/kg per mt in saline; IP; 30 min presession) and chlordiazepoxide HCI (Sigma; l, 3, 5, 8 mg/kg per ml in saline; IP, 30 min presession). Scopolamine- MBr was not tested because of evidence that does not support the assumption that the effects of the quaternary derivative are exclusively due to receptor blockade in the periphery (Moore et al. 1992b). Two doses of a third experimental drug, MDL 26,479, were also tested (data are not presented). Drug and dose sequences were counterbalanced across the six animals; in addition, the modality of the first unimodaI block was counterbalanced across the drug doses. Drug tests were separated by two vehicle sessions or as many vehicle sessions as were required for the individual animal to regain baseline performance (see above for criteria).

Behavioral measures and statistical analysis. For the final task, the following measures were recorded per stimulus type (four possible modality-quality pairings) and per block in each session: percentage of correct responses (i.e., number of correct responses divided by number of correct plus incorrect responses), percentage of the number of completed trials (number of correct plus incorrect responses

divided by the total number of trials), and correct-response latencies (time between onset of stimulus and correct lever response). As percentage data usually are not normally distributed, they were angularly transformed (arcsine of square root; Zar 1974). For the analysis of performance in the unimodal blocks (2 blocks per 20 trials), the first five trials of each block were discarded to exclude performance of the initial five trials of block 1 and the effects of the first shift from one to the other modality (first five trials of block 2). This was done to obtain measures of performance in unimodal blocks of trims that are relatively unaffected by task-onset and a shift in modality. The effects on the behavioral variables were collapsed across modality and quality and analyzed using a two-way repeated measures ANOVA (condition (i.e., uni- versus bimodal) x dose). The effects of CDP and scopolamine were analyzed indepen- dently. Significant effects were further analyzed using one-way repeated measures ANOVAs and paired t-tests (~ = 0.05). One animal discontinued responding during the bimodal condition following the highest dose of scopolamine. To include this animal's data in the statistical analysis, the two missing values were replaced by the mean of the other animals' data in this category.

Separate analyses were conducted in order to determine the effects of condition and dose on the animals' responses to visual versus auditory stimuli (modality) or to flashing/pulsing versus constant stimuli (quality; three-way ANOVAs: condition x modality or quality x dose). As a result of the highest doses of scopolamine and CDP, no responses for some stimulus categories and animals were obtained (e.g., no responses to pulsing tone). Therefore, the effects of the highest doses of both drugs were not included in this analysis. One remaining missing value was replaced by the corre- sponding value from the other modality, thus reducing the likeli- hood that effects of modality were statistically detected.

Additional post-hoc analyses addressed the possibility that response latency and accuracy changed across trials in a daily session by comparing the baseline performance during the first 30 sessions of the bimodal condition with the second half of trials in this condition. Statistical analyses were computed using the SPSS/PC + 5.0.1 software (SPSS Inc., Chicago, Ill.).

Results

Tab le 1 p r o v i d e s a s u m m a r y of the m a i n s ta t i s t ica l resul t s f o u n d in this e x p e r i m e n t .

Response latency

T h e ana lys i s o f the effects o f s c o p o l a m i n e r e v e a l e d signif- i can t m a i n effects o f c o n d i t i o n ( u n i m o d a l ve r sus b i m o d a l ;

Table 1. Synopsis of main statistical results

Measure Condition Dose C x D (C) (D)

Scopolamine Response latency bi in * Response accuracy ns ns ns Completed trials ns de ns

Chlordiazepoxide Response latency bi(t) in ns Response accuracy ns ns ns Completed trials ns de ns

Condition, uni-vs bimodal; bi, increased in the bimodal condition when compared with the unimodal condition; bi(t), trend (P=0.055); in, dose-dependent increase; de, dose-dependent decrease; *, significant interaction (see text)

216

3.2

" ~ 3.0 o) m

2.8

_~ 2 .6

2.4

2.2

2.0

scopolamine ehlordiazepoxide

Fig. 2. Response latencies (mean +_ SEM) in the focused (unimodal) and the divided (bimodal) attention condition (data derived from the ANOVAs on the effects of the two drugs). The effects of condition reached significance in the analysis of the effects of scopolamine and revealed a trend (P = 0.055) in the analysis of the effects of chlordiazepoxide. Clear columns, focused condition; hatched columns, divided condition

see Fig. 2) and dose on response latencies [condition: F(1,5) = 32.21; P < 0.01; dose: F(3,15) = 8.25; P < 0.01]. Multiple comparisons indicated that the response latencies in the bimodal condition were significantly longer than in the unimodal condition [t(5) = 5.68; P < 0.01]. Response latencies were significantly longer following the administration of 0.06 mg/kg and 0.10mg/kg scopolamine [ t (5)= 3.46; P < 0.02; t(5) = 22.87; P < 0.001; see Fig. 3]. The effects of condition and dose interacted significantly [F(3,15)= 4.28; P < 0.023]. This interaction was a result of greater effects of all doses of scopolamine in the bimodal condition (all paired t-tests: P < 0.05; see Fig. 3).

Similar to the effects of scopolamine on response tatencies, administration of chlordiazepoxide (CDP) resulted in an increase in response latencies [F(4,20) = 3.31; P < 0.05]. As illustrated in Fig. 3, administration of the smallest dose of chlordiazepoxide resulted in shorter response latencies [t (5) = 3.36; P < 0.05], and higher doses dose-dependently increased response latencies [multiple comparisons only showed a trend for the highest dose; t (5 )= 2.27; P = 0.072; Fig. 3] when compared with the effects of vehicle. The analysis of the effect of condition did not reach significance but showed a trend for increased response latencies in the bimodal condition IF(l,5) = 6.23; P = 0.055; see Fig. 2]. The effects of the

drug and condition on response latencies did not interact significantly [F(4,20) = 1.00; P > 0.05].

Response accuracy

Neither Drug [CDP: F(4,20) = 0.29; P > 0.05; scopolamine: F(3,15) = 0.84; P > 0.05] nor Condition (both analyses: P > 0.05] affected response accuracy. Similar- ly, there were no significant interactions between the effects of dose and condition on response accuracy (all P > 0.05).

Number of completed trials

Administration of either drug decreased the number of completed trials [i.e., correct plus incorrect responses; scopolamine: F(3,15) = 5.91; P < 0.01; CDP: F(4,20) = 5.69; P < 0.01]. There was neither an effect of condition (unimodal versus bimodal; both analyses: P > 0.05), nor did the effects of the drugs interact with the effects of condition (all P > 0.05; see Fig. 4). Multiple comparisons indicated that the effects of dose were based on the decrease in the number of completed trials that resulted from the administration of the highest dose of scopolamine [t(5) = 3.67; P < 0.05] or chlordiazepoxide [trend; t(5) = 2.33; P = 0.067].

Post hoc analyses

The hypothesis that the increase in response latencies in the bimodal condition (i.e., the effect of "condition"; see above) was confounded by an effect of prolonged behavioral testing was examined (see Methods for the reasons why the sequence of unimodat /b imodal condition could not be varied). The results of a comparison between baseline response latencies of the first 30 trials and the second 30 trials of the bimodal session [F(1,5) = 0.22; P > 0.05] did not support the possibility the latencies increased in the course of testing. There was, however, an effect on response accuracy [F(1,5) = 13.02; P < 0.02] that was based on a small but significant increase in response accuracy from the first to the second half of the bimodal condition (from an average of 86 to 90% correct responses). Thus, these post hoc analyses did not suggest a performance decrement as a function of prolonged testing.

4.0

3.5

~.~ o 0.0

2.5

t~ 0 E.O

1.5

1.0

4.0

3.0

i 1.O

tq 1.5

veh/ele 0.03 0.08 O.l m g / k g s c o p o l a m i n e

1.0 vehicle t 3 5 8

m g / k g c h l o r d i a z e p o x i d e

Fig. 3. Effects of scopolamine and chlordiazepoxide on response latencies (mean _+ SEM). Both drugs produced a dose-dependent increase in response latencies, and the effects of scopolamine were significantly greater in the bimodal (divided) condition. O, unimodal; 0, bimodal

0 c)

99.9 - -

90 - -

80 - -

70 - -

6 0 - -

5 0 - -

4 0

3 0 - -

2 0 - -

I 0 - -

vehicle dose 1 dose 2 dose 3 dose 4

Fig. 4. Effects of scopolamine and chlordiazepoxide on the relative number of completed trials (correct plus incorrect responses divided by the total number of trials; mean + SEM). As there was neither an effect of condition (unimodal versus bimodal) nor a significant interaction between condition and dose, data are shown collapsed across condition. Note that, generally, the number of errors of omis- sion (the inverse of the number of completed trials) remained relatively small ( < 10%). The highest dose of either drug significantly decreased the number of completed trials. The nonlinear ordinate is a result of an angular transformation of the relative number of completed trials (see Methods). O, chlordiazepoxide; @, scopolamine

As there were no main effects of drug and condition on response accuracy, post-hoe analyses on stimulus modality and quality were limited to response latency. These analyses did not reveal systematic differences in response latencies to stimuli of different quality (i.e., pulsing/flashing versus constant). There was a significant three-way interaction [condition x stimulus quality x dose (scopolamine); F(2,10) = 6.26; P < 0.05]; however, further analysis failed to locate the source for this interaction. Similarly, the analysis of the effects of scopolamine did not reveal any significant effects of modality (visual versus auditory). However, the analysis of the effects of chlordiazepoxide disclosed a significant interaction between Condition and Modality [F(1,5) = 10.31; P < 0.05] which appeared to be based on a decrease in response latencies to tones from the unimodal to the bimodal condition (means; unimodal: 2.13 s; bimodal: 2.00 s) while latencies to visual stimuli increased (unimodal: 1.78 s; bimodal: 2.05 s; both differences failed to reach significance). There was also an interaction between Modality and Dose [F(3,15)= 6.22; P < 0.01] which, on the basis of multiple comparisons, appeared to reflect the decrease in latencies as a result of the administration of the lowest dose of the drug in response to auditory stimuli only (these interactions are also reflected in Fig. 6, which shows the divided attention costs for the stimuli of the two modalities; see below).

The fact that there were no drug effects on response accuracy, but on response latency, prompted a final analysis to examine the possibility that these findings are indicative for a speed-accuracy tradeoff. As illustrated by the upper part of Fig. 5, in the bimodal condition, both drugs dose-dependently increased response latencies while response accuracy remained largely unchanged.

3.0

217

o o

O

2.0

o ~ _ ~ o . . . ~ L : . ~ ............... °--'~"e" ............ 0

,.o ~'.o ~:~ 3:o ~:5 g.o r e s p o n s e l a t ency (sec)

o

I

t

0.8

0.4 ......17~O J

O j

00 _J .................. ...........................

-0.4 ] ~ 0

-0.8 0'.2 016 1;0 114 r e s p o n s e latency: bi - u n i m o d a l (see)

Fig. 5. Evidence for a speed-accuracy tradeoff in animals treated with chlordiazepoxide or scopolamine (cost-benefit analysis; Kinchla 1992). The upper figure shows the effects of both drugs on response latency plotted against response accuracy (using the angularly transformed percentage data) in the bimodal condition. Ar- rows point from the average effects of vehicle (lst point of either drug) to the effects produced by ascending order of doses (CDP: 1, 3, 5, 8 mg/kg; scopolamine: 0.03, 0.06, 0.1 mg/kg; the animal that terminated responding as a result of the highest dose of scopolamine (see Methods) was not included in this post-hoe descriptive analysis). Animals maintained response accuracy while response latencies generally increased dose dependently. To illustrate that the increase in response latencies was not simply a main effect of the drugs or unrelated to demands on processing, the lower graph shows the means of the individual differences between bimodal and unimodal response tatencies and response accuracies, tf the drugs' effects on response latencies were independent from the animals' efforts to maintain response accuracy, these effect should not differ between the uni- and bimodal condition. However, the differences in response tatencies are positive and largely dose dependent. As reflected by a lack of effect of condition and of dose on response accuracy, the differences between individual bi- and unimodal accuracy values are small (Similar to the upper graph, arrows point to the effects of ascending drug doses.) O, chtordiazepoxide; 0 , scopolamine

These data do not exclude the possibility that the effect on response latencies and the lack of effect on response accuracy were unrelated. Assuming that the animals maintained stable accuracy at the cost of response latencies, it was predicted that such a speed-accuracy tradeoff will be augmented by an increase in processing toad. In other words, if the increases in response latency were simply an effect of the drug, no differences between the drug-induced response latencies of the unimodal and bimodal condition would be expected. However, the lower

218

Table 2. Divided attention costs (in ms) as a function of drug and dose

Vehicle Dose 1 Dose 2 Dose 3 Dose 4

Chlordiazepoxide 38 - 70 108 175 803 Scopolamine 85 798 582 1204 n/a

n/a, not applicable; data indicate the means of the individual differences between bimodal and unimodal response latencies

1500

1200

900

600

300

0

- 3 0 0

- 6 0 0

i / " "J .

i f ~ i i

veh ic l e 1 2 3 4

d o s e s

Fig. 6. Mean divided attention costs (in ms) as a function of drug, dose, and modality. Doses (1,2,3,4) refer to doses of CDP (1, 3, 5, 8 mg/kg) and scopolamine (0.03, 0.06, 0.1 mg/kg). Note that, under vehicle conditions, divided attention costs occurred exclusively in response to visual stimuli. The non-linear dose-response curve for the scopolamine-induced divided attention costs for visual stimuli is also reflected in the cost-benefit analysis of Fig. 5 (lower graph). Scopolamine generally was more potent than CDP in increasing the divided attention costs in terms of latencies, as was also reflected by a significant interaction between condition and dose (see Discus- sion). C), CDP - visual; 0 , CDP - auditory; Vt, SCOP - visual; m, SCOP auditory

graph of Fig. 5 illustrated that the drugs' effects on response latency were greater in the bimodal than the unimodal condition, while the difference in accuracy remained relatively small. The "absolute divided attention costs" (i.e., the increase in response latency from the focused to the divided attention condition; Salthouse 1985) are listed in Table 2.

As an interaction between modality and the effects of condition and chlordiazepoxide on response latencies was observed (see above), Fig. 6 illustrates the divided attention costs selectively for the stimuli of the two modalities. In accordance with the significant interactions between modality and condition, at baseline condition, divided attention costs were largely based on costs that occurred in the processing of visual stimuli. Addi- tionally, scopolamine was more potent than CDP in increasing the differences between unimodal and bimodal response latencies to stimuli of both modalities.

Discuss ion

Animals were trained in a task designed to test crossmodal divided attention. In line with relevant findings in humans, and with our expectations, performance in the

divided attention condition primarily differed from that in the focused attention condition by an increase in response latencies, while response accuracy remained unaffected by condition. Similarly, the administration of chlordiazepoxide (CDP) or scopolamine did not affect response accuracy, but resulted in an increase in response latencies. The effects of scopolamine were greater in the bimodal than in the unimodal condition. Cost-benefit analyses provided evidence for the hypothesis that stable response accuracy was maintained at the cost of response latency. Further analyses revealed different divided attention costs for the processing of visual and auditory stimuli, and differential potencies of CDP and scopolamine to increase these costs. Finally, the number of completed trials decreased as a result of the administration of both drugs; as these effects (i.e., on the inverse of the number of errors of omissions) may result from various mechanisms that remain largely unknown, they will not be further discussed. Different aspects of validity of this task and the significance of the drug effects on response latency and accuracy will be evaluated below.

The expected lack of detrimental effects on response accuracy requires further comment (see also Introduc- tion). Performance of well-trained rats in conditional discrimination tasks was previously found relatively insensitive to the effects of basal forebrain lesions (Everitt et al. 1987), or of drugs including BZR ligands or cholinergic drugs (Sarter 1990; Andrews et al. 1992; Dudchenko and Sarter 1992). It was speculated that extensive training of this task results in habit-based performance. The obser- vation that animals well trained in a visual conditional discrimination task were extremely resistant in acquiring a reversal (Dudchenko and Sarter, unpublished observa- tion) supports such a view. The processing required for the execution of habits was described as "automatic", "uncontrolled", "effortless", "implicit", or "procedural" (Kesner 1984; Morris 1984). Although the hypothetical characterization of the processing underlying habits in cognitive terms is not expected to improve the behavioral analysis of performance in such tasks (see Hineline 1984), it allows the prediction that the accuracy of such behavioral responses are fairly insensitive to a wide range of behavioral or pharmacological manipulations. It is speculated that, in the present task, the division in attention occurred in retrieving the correct rule for the modality; however, once the rule was retrieved, the response was habitual. Furthermore, the animals were trained to a point at which rules could be retrieved with reliable accuracy. Thus, in the present experiment, response accuracy was not expected to represent a primary indicator of the increased processing load in the divided attention condition, or the effects of drugs. Similar considerations resulted in a focus on response times as the crucial dependent measure in various human studies (see Introduction).

The task used in the present experiment presumably required the animals to perceive visual and auditory stimuli, to successively discriminate between different stimuli within one modality, and to choose one of two response alternatives (left or right lever). Thus, in the divided attention condition (modality uncertainty), animals were assumed to divert their processing capacity between the parallel or successive processing of both sets

219

of response rules. The results from the present experiment in rats parallel some major findings of Hohnsbein's et al. experiment in humans (the comparison refers to their 'choice reaction time' version of the task, as it appears more akin to the response-choice component of our task than their simple choice reaction time version). In both experiments, performance in the divided (bimodal) attention condition, when compared with the focused (unimodal) condition, was characterized by stable response accuracy and increased response latencies [collapsed over modality by 43 ms for humans (Hohnsbein et al. 1991), and by approximately 62 ms for rats (average difference of vehicle effects of both analyses; see Table 2)]. In con- trast to the human data (see also McDowd and Craik 1988), the divided attention costs for rats appeared higher for visual than for auditory stimuli. Thus, with the exception of the direction of an asymmetric allocation of resources to the processing of stimuli of one modality, the effects of the divided attention condition in the animal experiments show considerable face validity in terms of modelling the effects found in humans.

At the most conservative level of interpretation, the present data indicate that, in comparison with the blocks of trials with stimuli from one modality, the condition of modality uncertainty was a more complex task for the animals (Salthouse 1982). It remains speculative whether the increase in response latencies is indicative of the limitations in the animals' processing capacity, or whether they regarded the four stimulus categories as completely independent conditioned stimuli (this statement appears to hold true for the human literature; Hirst 1986; Pashler 1993). However, if the animals performed on the basis of four independent conditioned stimuli, an increase in response latencies during the bimodal condition would not necessarily be expected. Additionally, drug effects indicative of increased speed-accuracy tradeoffs would not be expected (see above). Therefore, the increase in response latencies may have indicated the additional time required to shift attentional resources between the processing of two competing pairs of propositional response rules. The validity of this speculation hinges largely on the evidence for speed-accuracy tradeoffs that result from cost-benefit analyses (Fig. 5).

Both the administration of scopolamine and chlordiazepoxide increased the divided attention costs in terms of response latency. The costs for the processing of stimuli of both modalities were increased as a result of the administration of scopolamine (the reasons for the nonlinear dose-response curve for visual stimuli are unknown; see Fig. 6 and also the lower graph of Fig. 5). In comparison, the smallest dose of chlordiazepoxide decreased the costs by approximately 70 ms (see Table 2 and Fig. 5). Figure 6 suggests that this finding was primarily due to the effects on auditory stimuli [the post hoc statistical analyses on modality (see Results) also indicated that the effect of 1 mg/kg CDP on latencies to auditory stimuli differed from baseline latencies]. There was also a (non-significant) decrease in latency costs for visual stimuli (see above); therefore, the visual and auditory latency cost changes from vehicle to the effects of 1 mg/kg CDP were roughly comparable (visual: 139ms; auditory: 159 ms; see Fig. 6). These effects of the smallest dose of

CDP may be related to the loss of response inhibition that has been frequently observed following small doses of benzodiazepine receptor agonists (Margules and Stein 1967). Thus, the available evidence is inadequate to con- clude that CDP induced a dissociation in the divided attention costs between visual and auditory stimuli. Rather, the data appear to indicate that lower doses of CDP generally did not affect baseline latency costs in either modality. Thus, the effects of scopolamine and CDP appear to differ primarily in terms of potency.

At baseline conditions, however, divided attention costs differed between responses to auditory and visual stimuli (as indicated by a significant modality by condition interaction; see Results). In other words, when compared to the unimodal condition, the increased response latencies in the bimodal condition were based on the increased latencies to visual stimuli (see Fig. 6). The source for this difference remains unknown. Differences in stimulus discriminability would be expected to similarly affect performance in both conditions and thus would not increase the divided attention costs. The asymmetric allocation of processing resources in favor of the processing of auditory stimuli (which is opposite to the findings in humans; see above) may represent a species-specific properties of rats.

The validity of these interpretations is possibly limited by two circumstances. First, the absence of an effect of prolonged testing on response latencies does not exclude the possibility that the effects of the drugs and the sequence of blocks and trials interacted (see Methods and Results) in a manner unrelated to assumptions about the effects of divided attention. Second, although the descriptive cost-benefit analyses (see Figs 5 and 6) were based on individual differences between the performance in the two conditions, a significant interaction between the effects of chtordiazepoxide and condition on response latencies would have strengthened the basis for a cost-benefit analysis of the effects of this drug. However, these criticisms do not seem to invalidate the findings that response latencies increased under the condition of modality uncertainty, and that scopolamine's effects on response latencies interacted with condition, while response accuracy remained unchanged.

The differences between the effects of scopolamine and CDP may be described primarily as differences in potency. Aside from the potency differences, both drugs produced qualitatively similar effects on performance in this task. This conclusion appears in line with the interpretation of the effects of both muscarinic antagonists and benzodiazepine receptor agonists in human studies in terms of impairments in central executive functions (Cur- ran 1991b; Curran et al. 1991; Rusted et al. 1991). Final- ly, the present findings are in line with the hypothesis that benzodiazepine receptor ligands mediate attentional abilities through their effects on cortical acetylcholine (Sarter et al. 1992). This hypothesis predicts that the attentional effects of benzodiazepine receptor agonists match the effects of muscarinic antagonists and, more generally, that benzodiazepine receptor ligands modulate cortical acetylcholine release. Behavioral and neuro- chemical findings have indicated that the validity of this hypothesis is restricted to situations associated with in-

220

creases in ac t iv i ty in chol inerg ic neu rons ( D u d c h e n k o and Sar te r 1991, 1992; M o o r e et al. 1992c, 1993). A n i m a l t asks of a t t en t iona l abi l i t ies such as the presen t c ros smoda l d iv ided a t t en t ion task will be i m p o r t a n t to demon- s t ra te tha t ac t iv i ty changes in chol inergic systems corre- late wi th d e m a n d s on a t t en t iona l processing, and with the effects of benzod iazep ine recep to r l igands on performance.

Acknowledgements. This research was supported in part by the PHS Grants AG10173 and MH46869.

References

AIlport A (t989) Visual attention. In: Posner MI (ed) Foundations of cognitive science. MIT Press, Cambridge, pp 631-682

Andrews JS, Griitzner M, Stephens DN (1992) Effects of cholinergic and non-cholinergic drugs on visual discrimination and delayed visual discrimination performance in rats. Psychopharmacology 106:523-530

Baddeley AD (1992) Working memory. Science 255:556-559 Craik FIM, Byrd M (1982) Aging and cognitive deficits. The role of

attentional resources. In: Craik FIM, Trehub S (eds) Aging and cognitive processes. Plenum Press, New York, pp 191--211

Curran HV (1991a) Antianxiety treatments and human memory. In: Briley M, File SE (eds) New concepts in anxiety. CRC Press, Boca Raton, pp 395-405

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

Curran VA, Schifano F, Lader M (I991) Models of memory dys- function? A comparison of the effects of scopolamine and lorazepam on memory, psychomotor performance and mood. Psychopharmacology 103:83 90

Dudchenko P, Sarter M (1991) GABAergic control of basal forebrain cholinergic neurons and memory. Behav Brain Res 42:33- 41

Dudchenko P, Sarter M (1992) Failure of chlordiazepoxide to re- produce the behavioral effects of muscimol administered into the basal forebrain. Behav Brain Res 47:202-205

Dudchenko P, Paul B, Sarter M (1992) Dissociation between the effects of benzodiazepine receptor agonists on behavioral vigilance and responsivity. Psychopharmacology 109:203-211

Dunne MP, Hartley LR (1986) Scopolamine and the control of attention in humans. Psychopharmacology 89:94-97

Everitt BJ, Robbins TW, Evenden JL, Marston HM, Jones GH, Sirki/i TE (1987) The effects of excitotoxic lesions of the substan- tia innominata, ventral and dorsal gtobus pallidus on the acqui- sition and retention of a conditional visual discrimination: implications for cholinergic hypotheses of learning and memory. Neuroscience 22: 441-469

Falkenstein M, Hohnsbein J, Hoorman J, Blanke L (1991) Effects of crossmodal divided attention on late ERP components. II. Er- ror processing in choice reaction tasks. EEG Clin Neurophysiol 78:447-455

Filoteo JV, Dells DC, Massman PJ, Demadura T, Butters N, Salmon DP (1992) Directed and divided attention in Alzheimer's disease: impairment in shifting of attention to global and local stimuli. J Clin Exp Neuropsychol 14:871-883

Grady CL, Grimes AM, Patronas N, Sunderland T, Foster NL, Rapaport SI (1989) Divided attention, as measured by dichotic speech performance, in dementia of the Alzheimer type. Arch Neurol 46:317-320

Hineline PN (1984) Can a statement in cognitive terms be a behavior-analytic interpretation ? Behav Anal 7: 97-100

Hirst W (1986) The psychology of attention. In: LeDoux JE, Hirst W (eds) Mind and brain. Dialogues in cognitive neuroscience. Cambridge University Press, Cambridge, pp 105-141

Hohnsbein J, Falkenstein M, ttoorman J, Blanke L (t 991) Effects of crossmodal divided attention on late ERP components. I. Sim- ple and choice reaction time tasks. EEG Clin Neurophysiol 78:438-446

Kesner RP (1984) The neurobiology of memory: implicit and ex- plicit assumptions. In: Lynch G, McGaugh JL, Weinberger NM (eds) Neurobiology of learning and memory Guilford Press, N.Y., pp 111-118

Kinchla RA (1992) Attention. Annu Rev Psychol 43:711-742 Margules DL, Stein L (1967) Neuroleptics versus tranquilizers: evi-

dence from animal studies of mode and site of action. In: Brill H, Cole JO, Deniker P, Hippius t1, Bradley PB (eds) Neuropsy- chopharmacology. Excerpta Medica, Amsterdam, pp 108-120

McDowd JM, Craik FIM (1988) Effects of aging and task difficulty on divided attention performance. J Exp Psychol [Hum Percept] 14:267-280

Moore H, Dudchenko P, Bruno JP, Sarter M (1992a) Toward mod- eling age-related changes in attentional abilities in rats: simple and choice reaction time tasks and vigilance Neurobiol Aging 13:759-772

Moore H. Dudchenko P, Comer KS, Bruno JP, Sarter M (1992b) Central versus peripheral effects of muscarinic antagonists: the limitations of quaternary ammonium derivatives Psychophar- macology 108:241-243

Moore H, Sarter M, Bruno JP (1992c) Age-dependent modulation of in vivo cortical acetylcholine release by benzodiazepine receptor ligands. Brain Res 596:17-29

Moore H, Sarter M, Bruno JP (1993) Bidirectional modulation of stimulated cortical acetylcholine release by benzodiazepine receptor inverse agonists. Brain Res 627:267-274

Morris RGM (i984) Is the distinction between procedural and declarative memory useful with respect to animal model of am- nesia? In: Lynch G, McGaugh JL, Weinberger NM (eds) Neuro- biology of learning and memory. Guilford, N.Y., pp 119-I24

Olton DS, Wenk GL, Church RM, Meck WH (1988) Attention and the frontal cortex as examined by simultaneous temporal processing. Neuropsychology 26:307-318

Pashler H (1993) Doing two things at the same time. Am Sci 81:48- 55

Preston GC, Broks P, Traub M, Ward C, Poppleton P, Stahl SM (1988) Effects of lorazepam on memory, attention and sedation in man. Psych opharmacology 95:208-215

Rusted JM, Eaton-Williams P, Warburton DM (1991) A comparison of the effects of scopolamine and diazepam on working memory. Psychopharmacology 105:442-445

Salthouse TA (1982) Adult cognition: an experimental psychology of human aging. Springer, New York

Salthouse TA (1985) Speed of behavior and its implications for cognition. In: Birren JE, Schaie KW (eds) Handbook of the psychology of aging. Van Nostrand Reinhold, New York, pp 400-426

Sarter M (1990) Retrieval of well-learned propositional rules: insensitive to changes in activity of individual neurotransmitter systems? Psychobiology 18:451-459

Sarter M, Dudchenko P, Moore H, Holley LA, Bruno JP (1992) Cognition enhancement based on GABA-cholinergic interactions. In: Levin ED, Decker M, Butcher LL (eds) Neurotrans- mitter interactions and cognitive function. Birkhfiuser, Boston, pp 329-354

Schneider W, Shiffrin RM (1977) Controlled and automatic human information processing. I. Detection, search and attention. Psy- chol Rev 84:1-66

Wesnes K, Simpson P, Kidd A (1988) An investigation of the range of cognitive impairments induced by scopolamine 0.6 mg s.c. Hum Psychopharmacology 3: 27-4t

Wickens CD (1984) Engineering psychology and human performance. Harper Collins, Columbus, Ohio

Wright RE (1981) Aging, divided attention, and processing capacity. J Gerontol 36:605-614

Zar JH (1974) Biostatistical analysis. Prentice Hall, N.J.

Crossmodal divided attention in rats: effects of chlordiazepoxide and scopolamine

Documents

Transcript of Crossmodal divided attention in rats: effects of chlordiazepoxide and scopolamine