Behavioural Brahz Research, 53 (1993) 11-20 �9 1993 Elsevier Science Publishers B.V. All rights reserved. 0166-4328/93/$06.00 1 1

BBR 01392

Three independent factors characterize spontaneous rat motor activity

M a r t i n P. P a u l u s '' 'b a n d M a r k A. G e y e r b

a Laboratory of Biological Dynamics and Theoretical Medicine, bDepartment of Psychiatry, School of Medicine, University of California San Diego, La Jolla, CA 92093 (USA)

(Received 12 May 1992) (Revised version received 4 September 1992)

(Accepted 6 October 1992)

Key words: Locomotor activity; Spatial scaling exponent; Temporal scaling exponent; Dynanfical entropy; Exploratory behavior; Factor analysis

A scaling apprqach was introduced recently to assess sequential and geometrical aspects of animal behavior. This study describes the rela- tionship between different measures of unconditioned motor activity of rats. Specifically, the amount of motor activity was assessed using both a traditional photobeam break measure, counts, and the temporal scaling exponent, cq which describes the ratio of fast to slow behavioral micro-events. The sequential characteristics of the behavior were assessed by the dynamical entropy, h, describing the degree of unpredictability of future movements. The geometrical characteristics of rat motor activity were quantified by the spatial scaling exponent, d. Exploratory ac- tivity was measured by counting rearings and holepoking responses. A factor analysis of these measures was conducted based on results from 137 drug-naive animals that were tested for I h in the Behavioral Pattern Monitor. Three independent factors account for 77~ of the variance. These factors can be described as the 'amount of activity', 'sequential response organization', and 'exploratory activity'. The factor loadings support the initial hypothesis that the geometrical structure of rat motor activity, i.e. the spatial scaling exponent d, varies independently from the amount of activity, i.e. counts. In addition, the distribution of these measures did not deviate significantly from normality suggesting that the z-scores of these variables, which have been used previously in the d-c~ plane description, are able to indicate significant changes of behavior. These results suggest that unconditioned motor activity is influenced by at least three independent factors. The independent assessment of these factors may contribute significantly to the understanding of the neural substrates involved in the organization of unconditioned behavior.

INTRODUCTION

Among the important screening procedures used routinely in animal psychopharmacology, uncondi- tioned motor activity paradigms have been described as the least standardized, most individualized and most vulnerable to environmental factors 4L. Nevertheless, unconditioned motor activity is used frequently to dis- tinguish drug effects, categorize different pharmacolog- ical substances L3, or assess behavioral correlates of neurobiological manipulations. The discrepancy be- tween the difficulty to extract sensitive, reliable, and meaningful quantitative descriptors and the wide ap- plicability of this paradigm may not necessarily be al- leviated by the progression of technology. While the use of computerized experimental equipment allows the ex- traction of an increasing number of quantitative de-

Correspondence: M.P. Paulus, Department of Psychiatry 0804, Uni- versity of California San Diego, La Jolla, CA 92093-0804, USA. Fax: (1)(619)543-2493.

scriptors of motor behavior tt'2~176 multiple, non- independent, statistical comparisons of these descriptors may yield results that are difficult to inter- pret due to the accumulation of significant effects by chance 37. The use of multivariate methods that extract latent variables is one approach to overcome this prob- lena. However, even though these variables may dis- criminate the effects of drugs and neurobiological ma- nipulations with considerable sensitivity 38, the behavioral significance of latent variables is often dif- ficult to specify.

Moreover, this approach exposes a rather deep con- ceptual problem underlying the difference between observer-based and automated assessments of motor activity. For the latter, individual behavioral responses having a-priori definitions dictated by the particular instrument are counted and cunmlated over selected time intervals. Computerization has encouraged the creation of multiple definitions of these behavioral re- sponses, despite the fact that they frequently assess a common dimension of motor activity. For example, the

12

a-priori definition and independent assessment of movements based on different distances between pho- tobeams or the velocities of movements based on these photobeam arrangements can be used to obtain multi- ple quantitative descriptors. However, it is often the case that these multiple measures are highly inter- correlatcd and therefore largely redundant. More im- portantly, while the approach of counting and cumu- lating discrete events may be appropriate for simple aspects of unconditioned motor behavior, it does not assess the presumably purposeful organization of be- havioral elements into sequences of behavior. By con- trast, an observer using rating scales commonly records characteristic combinations of individual elements of the animal's behavior. For example, it has been noted for rats that grooming behavior consists of a complex sequence of individual acts, whose topography can be affected sensitively by drugs zz. Thus, in order to un- derstand the behavioral organization of individual be- havioral elements a different approach is necessary to extract quantitative descriptors from the automated equipment.

Recently, we developed two different approaches based on the ergodic theory of non-linear dynamical systems 27 and fractal geometry 28 to assess quantita- tively the dynamical and geometrical aspects of motor response sequences. In both approaches, we assumed that a functional relationship exists between a variable of interest (e.g. a movement) and an experimental pa- rameter (e.g. photobeam arrangement). The extraction of the exponent quantifying this functional relationship provides a usefifl quantitative descriptor assessing fun- damentally different properties of unconditioned motor behavior. It was assumed initially that the geometrical and dynamical aspects of behavioral sequences vary independently from the amount of motor activity. Sub- sequcntly, this assumption was supported by the obser- vation that various drugs differentially affected the geo- metrical or dynamical characteristics vs. the amount of locomotor behavior. That is, drugs were found to alter the spatial and temporal patterns of locomotion with- out necessarily changing the amount of locomotion, and vice v e r s a 9't~ However, these findings do not address the question as to whether these measures re- flect normally independent dimensions of uncondi- tioned motor behavior. Specifically, we lacked evidence that these measures constitute different innate dimen- sions in naive undrugged animals. Hence, the aim of the current investigation was to determine whether uncon- ditioned motor activity can be decomposed into inde- pendent factors describing inherently different charac- teristics of unconditioned motor behavior.

In previous investigations, some investigators found

that while 'walks' and 'rears' did not correlate signifi- cantly, these variables accounted for 85~ of the vari- ance of motor behavior in standard photocell activity cages '7 suggesting that activity is not a unitary variable. Others 42 have shown that motor behavior in an open field environment is factorially complex. Specifically, increases of defecation covaricd independently from amount of activity allowing the extraction of two inde- pendent factors that were labelled 'emotional reactivity' and 'exploration'. Exploratory behavior has also been distinguished theoretically from motor activity and its function has been described as a transformation of novelty into familiarity 23. Lastly, drugs have also been reported to affect differentially the amount of motor activity and exploratory activity. Specifically, D-am- phetamine was found to stimulate motor activity but decrease exploration 3t. These investigations supply ample evidence that unconditioned motor behavior of the rat reflects at least two measurably distinct and independent components, namely the amount of motor activity and exploratory activity 21"3~'42. It was hypoth- esized that, in addition to these two factors, a third factor exists that describes the geometrical structure of motor activity. This subdivision had been conceptual- izcd pre~,iously by the description of drug effects in the d-ct plane, a two-dimensional z-score transform of the spatial and temporal scaling exponents, respectively 2s. Other pharmacological studies have shown that drugs affecting the dopaminergic system can disrupt the se- quential organization of behavioral responses 2~. In fact, D-amphetamine's effects have been conceptualized as a progressive disruption of complex behavioral catego- ries in favor of increasingly stereotyped behavior 22. These results would favor the hypothesis that measures of the sequential organization of responses may con- stitute independent and important determinants of un- conditioned motor activity.

MATERIALS AND METHODS

Several groups, adding up to 137, of male Sprague- Dawley rats weighing 250-300 g (Harlan, San Diego, CA or Madison, WI) that were kept on a reverse 12-h dark-light cycle were pooled for this evaluation. While the results of the pharmacological experiments have been reported elsewhere 3"4'28, the recorded data from these animals were re-used to investigate the underly- ing factors of normal locomotor and exploratory be- havior. In all cases, rats were brought to the laboratory 1 h prior to testing and each animal was tested only once in the BPM.

The detailed description of the measures and the

assumptions inherent in the scaling approach can be found in refs. 27-30. Similarly, a detailed description of the testing apparatus is given in ref. 9. Briefly, the BPM consists of a 30.5 x 61.0 x 38.0 cm black Plexi- glas holeboard equipped with three floor holes, three wall holes on either side of the long walls of the cham- ber, and one wall hole on the back wall. Holepokes are recorded via infrared photobeams located in each hole. In addition, rearing against the wall is detected by a touchplate.mounted 15 cm above the floor. The loca- tion of the rat in the BPM chamber is sampled with a frequency of 18 Hz by an array of 4 x 8 infrared pho- tobeams located 2 cm above the floor. The status of these photobeams is used to obtain 7 x 15 different po- sitions with a resolution of 4 cm in the following man- ner. For both x and y photobeams, the position is computed by

8 4

Y" 2 i l l ( 0 - 1 Y~ 2 j H ( j ) - 1 i = l j = l

x = and y = N (H(i)) N (H(j))

Here H(/) or H(j) indicated simply whether a particu- lar photobcam is broken (H(0 -- I) or not (H(i) = 0) and N (H(i)) or N (H(j)) counts the number of photobeams broken. For example, the photobeam status '01 I0' for the ),-direction yield the ),-position:

O x l + l x 3 + I x 5 + O x 7

= 4 .

Using this scheme, it is possible to extract the maxi- mum information about the position of the animal in the BPM in the presence of simultaneously broken pho- tobeams.

The assessment of motor activity was based on the notion of a micro-event, i.e. the smallest instrument- resolvable change in the (rear/poke, x, y, thne) event space. The different aspects of motor activity were mea- sured by the number of micro-event counts during the 60 min in the BPM, the temporal scaling exponent, c~, the spatial scaling exponent, d, the dynamical entropy, h, the % rears~counts, and the %pokes/counts measure. Specifically, activity levels were assessed via two ap- proaches. First, the number of micro-events during the exposure in the BPM chamber corresponds to a tradi- tional activity measure of photobeam interruptions. Second, the analysis of the frequency of micro-events for a given duration, t, yielded a power law distribution of the form

N (t, At) ~ t - ~

13

which was described quantitatively by the exponent, c~. This exponent describes the relative contribution or ratio of short duration micro-events to long duration micro-events. The temporal scaling exponent a pro- vides a measure for the local 'degree of acting' and may be viewed as analogous to an inter-response time anal- ysis in schedule-induced behaviors.

Two presumably different aspects of the structure of motor activity were evaluated using the dynamical en- tropy, h, and the spatial scaling exponent, d. Concep- tually, the dynamical entropy is based on the assess- ment of the uncertainty generated by chaotic non-linear dynamical systems. This quantity is calculated from a partition that divides a four-dimensional phase space (x location, y location, x velocity, y velocity) in 32 parti- tion elements (o);). Specifically, given a transition prob- ability from one partition element to another (P(ilj)=p(%.lo)i)) Shannon's (1948) entropy formula H(o)le)) = Y,.]p(o)i)P(ilj)log(P(ilj)), quantifies the oc- currence frequency of new and different sequences of partition elements. The number of new and different behavioral sequences with increasing sequence length is thought to describe the dynamical structure of rat movement sequences and is quantified by the dynam- ical entropy, h,

h = lim I l l

I t ! - - * OO

Thus, an animal that frequently displays the same movement sequences will yield a low entropy measure, whereas a rat that shows a large variety of different movement sequences will be characterized by a large entropy measure.

The spatial scaling exponent is based on the charac- terization of fractal objects by a non-integer dimension measure. Specifically, the scaling exponent d relates the path length of a sequence of micro-events in the BPM (L(k)) to the measuring resolution (k) with which the sequence is observed. This measure describes the geo- metrical characteristics of movements. The distances between k consecutive micro-events are computed and the dependence of the path lengths on k is functionally expressed in the scaling hypothesis.

L(k ),~ k -a

The spatial exponent, d, quantifies this dependence. A low d value corresponds to straight or directed move- ments, whereas a high d value indicates highly local or circumscribed movements.

The exploratory aspect of motor activity was assessed by the rccording of rearing against the wall

14

and holepoking. Both measures were corrected for the amount of activity by dividing the total number of rears and holepokes by coaltts and expressed in percentage values.

The statistical analysis was carried out on a SUN microcomputer using the BMDP 6 program package. The deviation from a normal distribution of the differ- ent measure distributions was assessed with the Sha- piro and Wilk's tV statistic 37. A principal component factor analysis with varimax orthogonal rotation of the factors was used to obtain the factor solution. The number of retained factors was determined according to the eigenvalue > 1.0 criterion. The significance of individual correlation coefficients was determined from tabulated r-values according to an/'-distribution with 120 degrees of freedom.

RESULTS

Distributional characteristics of BPM measures The descriptive statistical measures of the six varia-

bles assessing unconditioned motor behavior in the BPM chamber (Table I) indicate that particularly d and h are distributed tightly around their mean values. The Shapiro and Wilk's IV statistics for testing normality indicates that none of the measures deviate significantly from a normal distribution. However, the P-values for d, ~, %pokes/counts, and particularly %rears/counts indicate a trend towards a non-normal distribution. The graphical display of the measure distributions (Fig. 1) reveals a negatively skewed d distribution, a slightly positively skewed ct distribution, a ieptokur- totic %pokes/cotmts distribution, and a plathokurtotic % rears/cott l l ts distribution.

Behavioral differences of anhnals from different suppliers The one-way ANOVA (vendor) did not indicate sig-

nificant differences for cottnts, d, %rears/cotozts, or h. However, ct (F1.137=11.55, P<0.05) and %pokes/ c o a n t s ( F I , 1 3 7 = 14.24, P < 0.05) differed significantly be- tween the Harlan San Diego and Harlan Madison Vendor. Specifically, Harlan rats from the San Diego vendor exhibited a larger ratio of short duration versus long duration micro-events yielding a larger average cc value (1.61 + 0.009) compared to the Harlan rats from the Madison vendor (1.540 + 0.019). In addition, Harlan rats from the San Diego vendor exhibited more %pokes/counts (4.43 + 0.166)compared to the Harlan rats from the Madison vendor (3.09 + 0.311).

Factor analysis of BPM measures The correlation matrix between the measures re-

vealed several significant correlation coefficients (Table II). Specifically, the temporal scaling exponent, ct, was correlated highly with the number of micro-events dur- ing 60 min exposure in the BPM (r= 0.701, P<0.001). In addition, the dynamical characterization of se- quences of micro-events using the dynamical entropy, h, was correlated significantly with the path pattern characteristics using the spatial scaling exponent, d (r = 0.483, P < 0.001). The dynamical entropy appeared to be correlated slightly with cotolts, t~, and %rears/ COlOllS.

The factor analysis retrieved three factors explaining 77~ ofthe total variance generated by the six measures (Table III). The first factor loaded on the counts mea- sure and tile temporal scaling exponent, c~, with a slight contribution by the dynamical entropy, h. This factor can be identified readily as describing the 'amount of activity' of the animals in the BPM chamber. The sec-

TABLE I

Descriptive statistics for saline controls

This table shows the descriptive statistics for the six measures (co,rots, ct, d, h, %rears~counts, %pokes/counts) obtained from 137 saline control animals in the BPM chambers. Only unique mode values are displayed. The W-test and its corresponding P-values indicate the deviation from a normal distribution.

Mean 11 lode 111edian S.D. S.E.•L Min. illax. 1 V-test P

Co,mts 5540 ~t 1.592 d 1.618 h 1.870

% rears 2.243 r

%pokes 4.085 COlllltS

1.656 1.915

5588 1148 98 2739 8932 0.985 0.783 1.599 0.095 0.0081 1.281 1.813 0.9731 0.133 1.622 0.041 0.0035 1.508 1.712 0.9732 0.128 1.873 0.062 0.0053 1.653 2.010 0.984 0.722

2.353 1.119 0.095 0.0 5.21 0.969 0.06

4.033 1.687 0.145 0.753 10.485 0.973 0.151

15

30 30 [

25 25 ~ l ~ 20 ~ 20

.~ ~.15 ~ ~ . : s ~

10 r

5 5

0 0

counts alpha 3O 30

25 25

20 ~ 20

g 15 ~. 15

~o ~ ,o

5 5

0 0

spatial d entropy

30 30

25 25

20 ~ 2o

~5

t . i0 ~ i0

5 5

% rears / counts % pokes I counts

Fig. I. This figure displays the distr ibution o f counts, oq d, h, %rears/counts, %pokes/counts.

ond factor loaded on the spatial scaling exponent, d, the dynamical entropy, h, and to a minor extent on the %pokes/cotmts measure. Since this factor involved

mostly the measures that assess dynamical and geo- metrical characteristics of sequences of micro-events, it may be called 'sequential response organization'. The

T A B L E 11

Correlation matrix

This table gives the correlation matr ix for counts, or, d, h, %rears/counts, %pokes/counts. Statistically significant correlation coefficients with P < 0 . 0 1 are indicated by asterisks.

Counts d ~t h % rears %pokes

COZIlIIS COllHtS

Counts 1.000

d - 0.140 1.000

ct 0.701" - 0.105 1.000 h 0.269* 0.483* 0.201 1.000

% rears O. 126 0.084 O. 15 i 0.232* 1.000 c o u n t s

%pokes 0.088 - 0.058 0.150 - 0.045 0.256* COlllll$

1.000

16

TABLE I11

Rotated factor sohaion for saline controls

This table summarizes the rotated factor solution of the varimax, principal component exploratory factor analysis. Variable coefficients with loadings of <0.25 are omitted. The eigenvalue indicates the relative variance explained by the factors.

Factor 1 Factor 2 Factor 3

Counts 0.92 0.89

d 0.85 h 0.29 0.84

% rears

COllllls

%pokes 0.26

COllllls

Eigenvalue 1.80 1.53

0.82

0.75

1.26

third factor included both %rears/counts and %pokes/ co t tn l s and can thus be conceptualized as the 'explor- atory activity' of the animals in the BPM chamber. The eigenvalues of the three factors indicated that, while the 'amount of activity' explained most of the overall var- iance, both the 'sequential response organization' and the 'exploratory activity' contribute significantly to the description of the unconditioned motor behavior.

The distribution of individual cases in factor space (Fig. 2) did not suggest the existence of subgroups of animals sharing specific constellations of factor load- ings. Instead, all animals appeared to be distributed randomly within the factor space.

A resampling procedure was employed to test the reliability of the factors extracted. Specifically, 15 • [mtmber of independent factors] (n = 45) animals were selected randomly from the sample population (n-- 137). The correlation matrix was computed and a factor analysis with subsequent varimax orthogonal ro- tation was performed. If the number offactors extracted and the variable loadings agreed with the total sample, the sample was labelled 'confirmed' otherwise 'non- confirmed'. Nine out of i0 resampled factor analyses retrieved 3 independent factors and 8 of the variable Ioadings agreed completely with the factor analysis of the entire sample. In one case, only two factors ac- counting for 62 ~o of the variance were extracted. Thus, these results indicated that the factor structure was stable even when computed froth a substantially smaller sample.

Lastly, the factor analysis was carried out separately for animals from both the San Diego and Madison vendor to determine whether the differences in ~ and %pokes/counts would account for an artificial inflation

Distribution of 137 drug-naive rats in the three factor space

sequential response o r g a n i z a t ~ '

exploratory 21 "" " " " " " " e l Ice

activity:v(3) I " "

amount of activity: v ( 1 )0

F(1) = {counts, o]

1=(2) = {d, entropy}

F(3) = {% pokes/counts, % rears/counts }

Fig. 2. This figure shows the distribution of the individual saline control animals in factor space. The distribution of these points in-

dicates no clustering of cases.

of the correlation matrix and thus yield a spurious fac- tor structure. However, the factor analysis yielded the same factor loading pattern in both cases. Specifically, three factors were extracted for the San Diego and Madison vendor accounting for 75~o and 85~o of the variance. In both cases, the first factor loaded on the COltnIs measure and the temporal scaling exponent, ~, with a slight contribution of the dynamical entropy, h. The second factor loaded exclusively on the spatial scaling exponent, d and h, and the third factor loaded on both %rears/counts and %pokes/carrots. Conse- quently, while the Harlan rats from the two vendors differed on some measures obtained in the BPM, the factor structure was identical for rats from both ven- dors.

DISCUSSION

Any thorough assessment of the behavioral effects of a psychoactive agent or a neurobiological manipula- tion, such as lesions or genetic manipulations, should include information as to its effect on unconditioned

motor activity ~'13'3~. However, any single measure of locomotor activity will be confounded with other as- pects of behavior. Thus, it is important to determine the inherently independent dimensions affecting uncondi- tioned motor behavior. The three main results of the current investigation support the notion that multiple variables contribute independently to unconditioned motor behavior. First, at least three independent fac- tors characterize rat locomotor behavior. Specifically, the 'amount of activity', the 'sequential response orga- nization', and 'exploratory activity' vary independently for a large group of saline-treated animals in their first exposure to the BPM chamber. Second, the variables representing these factors do not differ significantly from a normal distribution, thus confirming that z-transform scores appropriately characterize the sig- nificance of experimental differences, as used previously to describe drug effects 28'3~ Third, there appears to be

I . ~

no clustering wahm the three-dimensional factor space, suggesting that there are no discrete 'states' or combi- nations of variables, but rather a continuum of varia- tion between the three factors.

In early studies, measures of unconditioned motor behavior were interpreted frequently in terms of hypo- thetical psychological concepts such as curiosity 2, emo- tionality 5, and fear zS"az. However, such concepts ap- pear to assign behavioral elements arbitrarily to categories defined a-prioriby the experimenter. Such categories may or may not correspond to the physio- logical systems influenced by drugs or other manipula- tions. Thus, rather than relying on a-priori definitions of behavioral categories, it seems more appropriate to assess empirically the extent to which experimentally obtained measures describe independent dimensions of behavior. Subsequently, one can then consider how these dimensions relate to observer-defined behavioral concepts. For example, numerous authors have inter- preted unconditioned motor behavior in terms of arousal, either implicitly or explicitly 5'24'25'33. Though principally derived from the state of activation of the reticular activating system zS, the concept of arousal has undergone substantial revisions in light of recent neu- rochemical and neuroanatomical evidence that several neurotransmitter systems are involved in its modula- tion. It has even been proposed that arousal is not a scalar concept but involves specific patterns of behav- ioral activation 15. Thus, it is important to obtain em- pirical evidence for independent dimensions of uncon- ditioned motor behavior.

While the three factors identified in the present ap- proach provide a comprehensive description of uncon- ditioned motor behavior, they require further elabora- tion to relate them to other concepts and measures that

17

have been used to characterize this behavior. First, the factor 'amount of activity' loaded on a traditional mea- sure of motor activity and on the temporal scaling ex- ponent, ct. This factor reflects the dimension that has been used traditionally to classify psychoactive agents as 'depressants' or 'psychostimulants '13. The close cor- relation between a global measure of motor activity, counts, and the distribution of a local measure of the degree of acting, ~, indicates that a shift in overall levels of motor activity is accompanied inherently by a cor- responding re-arrangement of the durations of individ- ual micro-events. Specifically, the current result does not indicate that a qualitative difference exists between long and short duration micro-events. While such a qualitative difference would be reflected in independent changes within the distribution of micro-events giving rise to variations of ~ that may not necessarily corre- late with the level of activity, this result suggests some overall regulation of the contribution of different dura- tion micro-events. The slight contribution of the dy- namical entropy, h, to this factor suggests that an in- creased activity level is accompanied by an increased number of different movement sequences. It is tempt- ing to speculate that normal control animals increase their response diversity by increasing their activity lev- els, since h quantifies to some extent the path variety reflecting the response diversity in the BPM and there appears to be a monotonically increasing relationship between activity levels and h. This possibility is partic- ularly intriguing because the effects of some psycho- stimulants, e.g. high doses of D-amphetamine z7' co- caine 29, or M D M A 27 is to reduce h while increasing the activity level.

Second, the 'sequential response organization' factor requires a thorough examination since it was derived from measures describing relationships between micro- events. This factor loaded primarily on the spatial scal- ing exponent, d, and the dynamical entropy, h. While the first measure assesses geometrical aspects of micro- event sequences, the latter characterizes the predict- ability or unpredictability of sequences of behavioral elements. Thus, saline control animals exhibit a corre- lation between the number of different partition element sequences (11) and the relative increase in local or me- andering movements at the expense of directed or straight movements (d). This finding suggests that, as the animals deviate from a directed straight path, they become more unpredictable, implying that animals move most predictably when going towards a point of interest. A similar conclusion has been reached on the basis of qualitative observations about excursions from a home cage into an open field and back in a free ex- ploration paradigm 1'8. A distinction has been made be-

18

tween the more meandering path on the outward bound part of the excursion and the straighter movement se- quences characteristic of the directed return to the home cage or home area s. Thus, this factor might reflect a different cognitive set underlying the sequential organi- zation of movements in the BPM chamber. Corre- spondingly, drugs that affect cognitive processing, e.g. scopolamine ~9, also dramatically decrease both the spatial scaling exponent, d, and the dynamical entropy, h 28.

Third, the factor 'exploratory activity' loaded on the activity-corrected measures for rearing against the touchplate and holepokes into the 10 holes in the BPM chamber, %rears/cotmts and %pokes/cotmts, respec- tively. This factor corresponds to an independent di- mension of unconditioned motor behavior that has also been described by other investigators 2L3~'42. The re- sults of this factor, however, suggest that the division into diversive exploration and inspective exploration 2 does not correspond to independently varying dimen- sions. Instead, rearing behavior as a measure for di- versive exploration and holepoking behavior indicating inspective exploration vary concurrently in saline- treated animals. On the other hand, this result is con- sistent with previous examples of manipulations affect- ing both rearing and holepoking independently of the effects on the amount of activity. For example, the ac- tivation of hippocampal/~-adrenergic receptors prefcr- entially increases the measures of exploration 7. Con- versely, M D M A and related serotonergic releasers simultaneously decrease holepokes and rearings while increasing the amount of activity ~2. The significance of exploratory behavior has been interpreted as a trans- formation of novelty to familiarity, since the animal appears to take a certain initiative in finding out more about its environment 2"23. However, while it has been reported that D-amphetamine differentially decreased exploratory behavior 3~, the same report found that ex- ploratory behavior in undrugged animals, when con- sidered as an observing response maintained over time, is not incompatible with locomotor activity. Specifi- cally, the more highly active animals tended to explore novel stimuli more than less active animals. In contrast, this investigation reveals that the percentage of animals engaging in 'exploratory activity' does vary independent from the level of activity.

The fact that the concept of 'arousal' may actually correspond to a vectorial state rather than a scalar quantity is particularly emphasized when considering the influence of stress on the 'amount of activity' and 'exploratory activity' as independent aspects of uncon- ditioned motor behavior. Stress has been identified by some theorists specifically with emotional arousal, while

others have used the term to refer to any significant change in the normal environmental input tS. A number of investigations have supported the independence of both aspects with respect to stress-induced changes in behavior. For example, prior stress reduces exploratory behavior directed at novel objects but did not affect general activity 34. Furthermore, isolation stress reduced exploratory holepoking, increased locomotor activity, and increased the preference of mice for the open arms of a plus-maze 14. Similarly, a lateral habenula lesion potentiated the effects of footshock stress on explor- atory behavior in an open field without augmenting of general motor function ~8. Thus, the current investiga- tion further supports the inherent independence of these aspects of unconditioned motor behavior, which may correspond to distinct neural substrates. In addition, these results substantiate the possibility of differential assessment of motor behavior within the context of the unconditioned motor paradigm, which has been criti- cized by Hughes 16 who suggested that 'while studies may be useful for determining drug effects on non- directed locomotion or generalized random activity, they contribute little to the understanding of such ef- fects on exploration per se because of the difficulty in distinguishing between exploratory activities initiated by and directed towards environmental stimuli, and non-specific activities motivated entirely by some inter- nal stimulus or state'.

To summarize, spontaneous or unconditioned motor behavior is characterized by at least three independent factors that may be correlated with complex interac- tions between various neurotransmitters, such as the dopaminergic, serotonergic, cholinergic, and noradren- ergic neurotransmitter systems. These transmitters have been implicated in non-specific functions determining the overall 'state' of the animal. While the specific mod- ulation of the behavior is still not well understood, they are thought to affect processes involving response ini- tiation, generation, selection, and ordering. However, the modulation of these processes using pharmacolog- ical or neurobiological manipulations can be assessed more precisely by the projection of the unconditioned motor behavior on the three factors extracted here. Thus, this assessment may yield further insights into the various neural mechanisms underlying different aspects of unconditioned motor activity. For example, differ- ential effects on independent factors may indicate that distinct mechanisms or sites of actions underlie the effects of different manipulations. Moreover, differen- tial changes of variables that normally covary within one factor may reflect a decoupling of concomitant be- havioral responses, which may be interpreted as a re- sponse disorganization. In conclusion, unconditioned

motor activity paradigms provide sensitive probes of the behavioral response topography which consists of at least three independent factors.

REFERENCES

1 Adams, L.M. and Gcyer, M.A., l.SD-induced alterations in lo- comotor patterns and exploration in rats, Psychopharmacology, 77 (1982) 179-185.

2 Berlyne, D.E., Cot~ict. Arousaland C, ttiosity, McGraw Hill, New York, 1960.

3 Callaway, C.W., Johnson, M.P., Gold, L.II., Nichols, D.E. and Geyer, M.A., Amphetamine derivatives induce locomotor hyper- activity by acting as indirect serotonin agonists, Psychopharma- cology, 104 (1991) 293-301.

4 Callaway, C.W., Wing, L.L. and Gcycr, M.A., Serotonin release contributes to the locomotor stimulant effects of 3,4-methylcne- dioxymcthamphetamine in rats, J. Pharmacol. Exp. Ther., 254 (1990) 456-464.

5 Denenberg, V.I!., Open field behaviour in the rat: what does it mean? In E. Tobach, (Ed.), ExperimentalApproaches to the Study of Emotional Behavior, Vol. 1, SuppL 1, Ankho International, New York, 1969, pp. 163-174.

6 Dixon, W.J., BMDP Biomedical Computer Programs, University of California Press, Los Angeles, 1990.

7 Flicker, C. and Geyer, M.A., Behavior during lfippocampal mi- croinfusions: 1. Norepinephrine and diversive exploration, Brahl Res. Rev., 4 (1982) 53.

8 Geyer, M.A., Variational and probabilistic aspects of exploratory behavior in space: four stimulant styles, Psychophannacol. Bull., 18 (1982) 48-51.

9 Geyer, M.A., Russo, P.V. and Masten, V.L., Multivariate assess- ment of locomotor behavior: pharmacological and behavioral analyses. Phannacol. Biochem. Behav., 25 (1986) 277-288.

10 Geyer, M.A., Russo, P.V., Segal, D.S. and Kuczenski, R., Effects of apomorphine and amphetamine on patterns of locomotor and investigatory behavior in rats, Phannacol. Biochem. Behav., 28 (1987) 393-399.

I 1 Geyer, M.A., Approaches to the characterization of drug effects on locomotor activity in rodents. In M.W. Adler and A. Cowan (Eds), Testing and Evahtation of Dntgs of Abuse, Wiley-Liss, New York, 1990, pp. 81-99.

12 Gold, L.H., Koob, G.F. and Geyer, M.A., Stimulant and hallu- cinogenic behavioral profiles of 3,4-methylenedioxymethamphet- amine and N-ethyl-3,4-methylenedioxyanlphetanline in rats, .L Phannacol. Exp. Ther., 247 (1988) 547-555.

13 Harvey, J.A., Behavioral pharmacology of central nervous stim- ulants, Neuropharmacolog), 26 (1987) 887-892.

14 Hilakivi, L.A., Ota, M. and Lister, R.G., Effect of isolation on brain monoamines and the behavior of mice in tests of explora- tion, locomotion, anxiety and behavioral 'despair', PhannacoL Biochem. Behav., 33 (1989) 371-374.

15 Itockey, R., Stress and the cognitive components of skilled per- formance. In V. Hamilton and D. Warburton (Eds.), Human Stress and Cognition, Wiley, Chichester, 1979, pp. 141-178.

16 Ilughes, R.N., Chlordiazepoxide modified exploration in rats, Psychopharmacologia, 24 (1972) 462-469.

17 Krsiak, M., Steinberg H. and Stolerman I.P., Uses and limita- tions of photocell activity cages for assessing effects of drugs, Psychopharmacologia, 17 (1970) 258-274.

18 Lee, E.H. and Huang, S.L., Role of lateral habcnula in the rcg-

19

ulation of exploratory behavior and its relationship to stress in rats, Behav. Brain Res., 30 (1988) 265-271.

19 l.evin, E.D., Castonguay, M. and Ellison, G.D., Effects of the nicotinic receptor blocker mecamylamine on radial-arm maze performance in rats, Behav. NeuroL BioL, 48 (1987) 206-212.

20 Ljungberg, T. and Ungerstcdt, U. A method for simultaneously recording of eight behavioral parameters related to monoamine ncurotransmission, PhannacoL Biochem. Behav., 8 (1977) 483- 489.

21 Ljungberg, T. and Enquist, M., Disruptive effects of low doses of D-amphetamine on the ability of rats to organize behaviour into functional sequences, Psychophannacolog); 93 (1987) 146-151.

22 Lyon, M. and Robbins, T.W., The action of central nervous system stimulant drugs: a general theory concerning amphetamine effects. In W. Essman, and L. Valzclli (Eds.), Current Develop- ments hz Psychopharmacology, Vol. 2, Spectrum, New York, 1977, pp. 79-163.

23 McReynolds, P., Exploratory behavior: a theoretical interpreta- tion, Psychol. Rep., 11 (1962) 311-318.

24 Montgomery, K.C., The relation between fear induced by novel stimulation and exploratory behavior, J. Comp. Physiol. PsychoL, 48 (1955) 254-260.

25 Moruzzi, G. and Magoun, H.W., Brain stem reticular formation and activation of the EEG., EEG Clin. NeurophysioL, 1 (1949) 455-473.

26 Nickolson, V.J., Detailed analysis of the effects of apomorphine and D-amphetamine on spontaneous locomotor behaviour of rats as measured in a TV-based automated open-field system, E,r J. Pharmacol., 72 (1981)45-56.

27 Paulus, M.P., Geyer, M.A., Gold, L.H. and Mandcll, A.J., Ap- plication of entropy measures derived from the ergodic theory of dynamical systems to rat locomotor behavior, Proc. Natl. Acad. ScL USA, 87 (1990) 723-727.

28 Paulus, M.P. and Geyer, M.A., A temporal and spatial scaling hypothesis for the behavioral effects of psychostimulants, Psy- chophamzacolog); 104 (1991) 6-16.

29 Paulus, M.P., Geyer, M.A. and Mandell, A.J., Statistical me- chanics of a neurobiological dynamical system: The spectrum of local entropies (S(~)) applied to cocaine-perturbed behavior, Physica A, 147 (1991) 567-577.

30 Paulus, M.P. and Geyer, M.A., The effects of MDMA and other methylenedioxy substituted phcnylalkylamines on the structure of rat locomotor activity, Ne,tropsychophannacology, 7 (1992) 15-31.

31 Robbins, T.W. and Iversen, S.D., A dissociation of the effects of D-amphetamine on locomotor activity and exploration in rats, Ps)'chophannacologia, 28 (1973) 155-164.

32 Robbins, T.W., A critique of the methods available for the mea- surement of spontaneous motor activity. In L.L. Iversen, S.D. Iversen and S.tt. Snydcr (Eds.), llandbook of Psychophannacol- ogy, Vo! 7, Plenum, New York, 1977, pp. 37-82.

33 Robbins, T.W. and Everitt, B.J., Psychopharmacological studies of arousal and attention. In S. Stahl, S.D. Iversen and E. Good- man (Eds.), Cognitive Neurochemisto', Oxford University Press, Oxford, 1990, p. 135.

34 Rosellini, R.A. and Widman, D.R., Prior exposure to stress re- duces the diversity of exploratory behavior of novel objects in the rat (Rattus norvegicus), J. Comp. PsychoL, 103 (1989) 339-346.

35 Sanberg, I'.R., Zoloty, S.A., Willis, R., Ticarich, C.D., Rhoads, K., Nagy R.P., Mitchell, S.G., Laforest, A.R., Jcnks, J.A., Hark- abus, L.J., Gurson, D.B., Finnefrock, J.A. and Bednarik, E.J. Digiscan activity: automated measurement of thigmotactic and stereotypic behavior in rats, Pharmacol. Biochem. Behav., 27 (1987) 569-572.

20

36 Shannon, C.E., A mathematical theory of communication, Bell Sys. TechnoL ,L, 27 (1948) 370-423 + 623-656.

37 Shapiro, S.S. and Wilk, M.B., An analysis of variance for nor- mality (complete samples), Biometrika, 52 (1965) 591-611.

38 Stahle, L., Pharmacological Studies on Behavioral Changes indztced b)' Dopamhle Agonists hi the Rat: A 3hdtivariate Approach, Thesis, Karolinska University, Stockholm, 1987.

39 Stable, L. and Ungerstedt, U., Yawning and suppression of ex- ploration in amphetamine-treated rats, incompatibility with the autoreceptor hypothesis, Psychophannacologk; 97 (1989) 553-560.

40 Stoff, D.M., Stauderman, K. and Wyatt, R.J.,The time and space machine: continuous measuremen.t, of drug-induced behavior pat- terns in the rat, Psychopharmacology, 80 (1983) 319-324.

41 Watzman, N., Barry, M., Kinnard, W.J. and Bucklcy, J.P., Com- parison of different photobeam arrangements for measuring spontaneous activity of mice, J. PharmacoL Sci., 55 (1966) 907- 909.

42 Whimbey, A.E. and Denenburg, V.H., Two independent behav- ioral dimensions in open field performance, J'. Comp. PhysioL PsychoL, 63 (1967) 500-504.