Journal of Experimental Psychology

Journal of

Experimental Psychology

VOL. 74, No. 1 MAY 1967



FAILURE TO ESCAPE TRAUMATIC SHOCK 1

2

MARTIN E: P. SELIGMAN AND STEVEN F. MAIER«

University of Pennsylvania



Dogs which had 1st learned to panel press in a harness in order

to escape shock subsequently showed normal acquisition of escape/

avoidance behavior in a shuttle box. In contrast, yoked, inescapable

shock in the harness produced profound interference with subsequent

escape responding in the shuttle box. Initial experience with escape

in the shuttle box led to enhanced panel pressing during inescapable

shock in the harness and prevented interference with later responding

in the shuttle box. Inescapable shock in the harness and failure to

escape in the shuttle box produced interference with escape responding

after a 7-day rest. These results were interpreted as supporting a

learned "helplessness" explanation of interference with escape re-

sponding: Ss failed to escape shock in the shuttle box following in-

escapable shock in the harness because they had learned that shock

termination was independent of responding.



Overmier and Seligman (1967) have shock in the shuttle box. They initi-

shown that the prior exposure of dogs ally show normal reactivity to shock,

to inescapable shock in a Pavlovian but after a few trials, they passively

harness reliably results in interfer- "accept" shock and fail to make escape

ence with subsequent escape/avoidance movements. Moreover, if an escape or

learning in a shuttle box. Typically, avoidance response does occur, it does

these dogs do not even escape from not reliably predict future escapes or

1

This research was supported by grants to avoidances, as it does in normal dogs.

R. L. Solomon from the National Science This pattern of effects is probably

Foundation (GB-2428) and the National In- not the result of incompatible skeletal

stitute of Mental Health (MH-04202). The responses reinforced during the in-

authors are grateful to R. L. Solomon, J. escapable shocks, because it can be

Aronfreed, J. Geer, H. Gleitman, F. Irwin,

D. Williams, and J. Wishner for their advice shown even when the inescapable

in the conduct and reporting of these ex- shocks are delivered while the dogs

periments. The authors also thank J. Bruce are paralyzed by curare. This be-

Overmier with whom Exp. I was begun. havior is also probably not the result

2

National Science Foundation predoctoral

fellow. of adaptation to shock, because it occurs

8

National Institute of Mental Health pre- even when escape/avoidance shocks are

doctoral fellow. intensified. However, the fact that in-

1

MARTIN E. P. SELIGMAN AND STEVEN F. MAIER



terference does not occur if 48 hr. structed so that 5's legs hung down below

elapse between exposure to inescap- its body through four holes. The 5's legs

were secured in this position, and S was

able shock in the harness and escape/ strapped into the hammock. In addition, S"s

avoidance training, suggests that the head was held in position by panels placed

phenomenon may be partially depend- on either side and a yoke between the panels

ent upon some other temporary proc- across 5's neck. The S could press the

ess. panels with its head. For the Escape group

pressing the panels terminated shock, while

Overmier and Seligman (1967) sug- for the "Yoked" control group, panel presses

gested that the degree of control over did not effect the preprogrammed shock.

shock allowed to the animal in the The shock source for this unit consisted of

harness may be an important deter- 500 v. ac transformer and a parallel voltage

divider, with the current applied through

minant of this interference effect. Ac- a fixed resistance of 20,000 ohms. The shock

cording to this hypothesis, if shock is was applied to 5 through brass plate elec-

terminated independently of S's re- trodes coated with commercial electrode

sponses during its initial experience paste and taped to the footpads of 5"s hind

feet. The shock intensity was 6.0 ma. Shock

with shock, interference with sub- presentations were controlled by automatic

sequent escape/avoidance responding relay circuitry located outside the cubicle.

should occur. If, however, ,9's re- Escape/avoidance training was conducted

sponses terminate shock during its in a two-way shuttle box with two black

initial experience with shock, normal compartments separated by an adjustable

barrier (described in Solomon & Wynne,

escape/avoidance responding should 1953). The barrier height was adjusted to

subsequently occur. Experiment I in- 5"s shoulder height. Each shuttle-box com-

vestigates the effects of escapable as partment was illuminated by two 50-w. and

compared with inescapable shock on one 74-w. lamps. The CS consisted of turn-

ing off the four 50-w. lamps. The US,

subsequent escape/avoidance respond- electric shock, was administered through the

ing. grid floor. A commutator shifted the polar-

ity of the grid bars four times per second.

EXPERIMENT I The shock was 550 v. ac applied through a

Method variable current limiting resistor in series

with 5. The shock was continually regu-

Subjects,—The 5s were 30 experimentally lated by E at 4.5 ma. Whenever S crossed

naive, mongrel dogs, 15-19 in. high at the the barrier, photocell beams were inter-

shoulder, and weighing between 25 and 29 rupted, a response was automatically re-

Ib. They were maintained on ad lib food corded, and the trial terminated. Latencies

and water in individual cages. Three dogs of barrier jumping were measured from CS

were discarded from the Escape group, two onset to the nearest .01 sec. by an electric

because they failed to learn to escape shock clock. Stimulus presentations and temporal

in the harness (see procedure), and one be- contingencies were controlled by automatic

cause of a procedural error. Three dogs relay circuitry in a nearby room.

were discarded from the "Yoked" control White masking noise at approximately

group, two because they were too small at 70-db. SPL was presented in both units.

the neck to be adequately restrained in Procedure.—The Escape group received

the harness; the third died during treatment. escape training in the harness. Sixty-four

This left 24 5s, eight in each group. unsignaled 6.0 ma. shocks were presented

Apparatus.—The apparatus was the same at a mean interval of 90 sec. (range, 60-

as that described in Overmier and Seligman 120 sec.). If the dog pressed either panel

(1967). It consisted of two distinctively with its head during shock, shock termi-

different units, one for escapable/inescapable nated. If the dog failed to press a panel

shock sessions and the other for escape/ during shock, shock terminated automatically

avoidance training. The unit in which 5s after 30 sec. Two dogs were discarded for

were exposed to escapable/inescapable shock failing to escape 18 of the last 20 shocks.*

consisted of a rubberized, cloth hammock

located inside a shielded, white, sound-at- * It might be argued that eliminating these

tenuating cubicle. The hammock was con- two dogs would bias the data. Thus naive

ESCAPE FAILURE WITH TRAUMATIC SHOCK



Twenty-four hours later dogs in the Es- TABLE 1

cape group were given 10 trials of escape/ INDEXES OF SHUTTLE Box ESCAPE/AVOID-

avoidance training in the shuttle box: 5" was ANCE RESPONDING: EXP. I

placed in the shuttle box and given 5 min.

to adapt before any treatment was begun. %Sa

Presentation of the CS began each trial. Falling to Mean No.

The CS-US interval was 10 sec. If S Mean Escape Failures

Group Latency Shock on to Escape

jumped the barrier during this interval, the (in sec.) 9 or More Shock'

CS terminated and no shock was presented. of the 10

Trials

Failure to jump the barrier during the CS-

US interval led to shock which remained on Escape 27.00 0 2.63

until 5 did jump the barrier. If no response Normal Control 25.93 12.5 2.25

occurred within 60 sec. after CS onset, the "Yoked ' Control 48.22 75 7.25

trial was automatically terminated and a 60-

sec. latency recorded. The average inter- • Out of 10 trials.

trial interval was 90 sec. with a range of 60-

120 sec. If 5" failed to cross the barrier on

all of the first five trials, it was removed,

ing latencies of panel pressing over

placed on the other side of the shuttle box, the course of the session (p = .008,

and training then continued. At the end of sign test, Trials 1-8 vs. Trials 57-64).

the tenth trial, 5 was removed from the Individual records revealed that each

shuttle box and returned to its home cage. 6" learned to escape shock by emitting

The Normal control group received only

10 escape/avoidance trials in the shuttle box a single, discrete panel press following

as described above. shock onset. The -Ss in the "Yoked"

The "Yoked" control group received the control group typically ceased panel

same exposure to shock in the harness as did pressing altogether after about 30

the Escape group, except that panel pressing

did not terminate shock. The duration of trials.

shock on any given trial was determined by Table 1 presents the mean latency

the mean duration of the corresponding trial of shuttle box responding, the mean

in the Escape group. Thus each 5" in the number of failures to escape shock, and

"Yoked" control group received a series of

shocks of decreasing duration totaling to the percentage of Ss which failed to

226 sec. escape nine or more of the 10 trials

Twenty-four hours later, 5s in the during escape/avoidance training in

"Yoked" control group received 10 escape/ the shuttle box for each group. The

avoidance trials in the shuttle box as de- "Yoked" control group showed marked

scribed for the Escape group. Seven days

later, those Ss in this group which showed interference with escape responding in

the interference effect received 10 more the shuttle box. It differed signifi-

trials in the shuttle box. cantly from the Escape group and from

5 the Normal control group on mean

Results latency and mean number of failures

The Escape group learned to panel to escape (in both cases, p • HARNESS TREATMENT escapable group showed significantly

more improvement than the Preescape

group, Mann-Whitney U test, U.= 15,

PREINESC.

p < .05. No significant differences

were found on difference scores for

any subsequent blocks of trials.

'*»»».....,.,MM poc cop Figure 2 presents the mean number

'* """^NO INESC.' of failures to escape shock for the three

groups across the four blocks of trials.

1 2 3 4 Analysis of variance revealed a signifi-

TRIALS (BLOCKS OF IO) cant overall effect of blocks, F (3, 72)

FIG. 1. Mean median latency of escape/ = 5.94, p < .01, and a significant

avoidance responding. (The position of the Groups X Blocks interaction, F (6,

arrow denotes whether the harness treat- 72) = 17.82, p < .01. Duncan's tests

ment occurred 24 hr. before the first or

second block of trials.) indicated that the No Pregroup showed

significantly more failures to escape

than the other two groups across the

for the mean of mean latencies. A

40 trials, both p < .05. The Preescape

small, transitory disruption of improve-

and the No Inescapable groups did not

ment in shuttle-box performance fol-

differ.

lowing inescapable shock in the harness

occurred in the Preescape group rela- Figure 3 presents the total number

tive to the No Inescapable group. of avoidance responses for the groups

Difference scores for latencies between across the blocks of trials. Only the

consecutive blocks of trials measure blocks effect was significant in the

improvement in performance. A com- overall analysis of variance, F (3, 72)

parison of the Preescape group with = 27.90, p < .01. No other effects

the No Inescapable group on the dif- were significant.

ference between the mean latency on Panel presses made in the harness

Trials 1-10 and the mean latency on during the inescapable shock session

Trials 11-20 revealed that the No In-









,M INESC

NO PRE









....PRE ESC.

-xHO INESC.





1 2 3 4

TRIALS (BLOCKS OF 10) TRIALS (BLOCKS OF IO)

FIG. 2. Mean number of failures to escape FIG. 3. Mean number of avoidances.

shock. (The position of the arrow denotes (The position of the arrow denotes whether

whether the harness treatment occurred 24 the harness treatment occurred 24 hr. before

hr. before the first or second block of trials.) the first or second block of trials.)

ESCAPE FAILURE WITH TRAUMATIC SHOCK



were counted. On either side of 5"'s No Pregroup. This impression was

head were panels which 5" could press ; borne out by statistical tests. The Pre-

panel pressing had no effect on the inescapable group showed significantly

shock, but merely indicated attempts to slower median latency of barrier jump-

respond and/or struggling in the har- ing than the Preescape and the No

ness. The Preescape group, having Inescapable groups across all 40 trials,

received 10 trials with escapable shock both p < .05, Duncan's test. The Pre-

in the shuttle box the previous day, inescapable group did not differ from

made more panel presses during the the No Pregroup. Similar results held

inescapable shock session than did the for the other indexes.

No Pregroup, the group for which Analysis of the panel press data

the inescapable shock in the harness showed that the Preinescapable group

was the first experimental treatment, made significantly fewer panel presses

Mann-Whitney U test, U = 9, p < .02. in the harness than the Preescape

Posterior control group.—Subse- group, Mann-Whitney U test, U = 14,

quent to this experiment, a control p < .05. The Preinescapable group

group was run to determine if the did not differ significantly from the No

escapability of shock in the shuttle box Pregroup, U = 26.

on Day 1 for the Preescape group was

responsible for its enhanced panel Discussion

pressing in the harness and lack of Three main findings emerged from

interference with responding in the Exp. II: (a) 5s (Preescape), which first

shuttle box. Or would the mere oc- received escapable shock in the shuttle

currence of inescapable shock for a box, then inescapable shock in the har-

free-moving animal in the shuttle box ness, did not react passively to subsequent

have produced these results? Nine shock in the shuttle box, as did 5s which

naive dogs received the following treat- either first received inescapable shock in

the shuttle box (Preinescapable) or no

ment: On Day 1, 5"s were placed in treatment prior to shock in the harness

shuttle box and given 10 trials as for (No Pre). (6) The Preescape group,

the Preescape and the No Inescapable having received experience with escapa-

groups. Unlike these groups, how- ble shock in the shuttle box, showed en-

ever, S's barrier jumping did not (ex- hanced panel pressing when exposed to

cept adventitiously) terminate the inescapable shock in the harness, relative

shock and CS, because trial durations to naive 5s given inescapable shock in

were programmed independently of ,S"s the harness. Such enhanced panel press-

behavior. The duration of each of the ing was specifically the result of the es-

capability of shock in the shuttle box:

10 trials for this Preinescapable group The Preinescapable group did not show

corresponded to the mean trial dura- enhanced panel pressing, (c) The inter-

tion for the Preescape and the No In- ference effect persisted for 40 trials.

escapable groups on that trial. On The 5s which have had prior experi-

Day 2, 5"s received 64 trials of inescap- ence with escapable shock in the shuttle

able shock in the harness. On Day 3, box showed more energetic behavior in

Ss received 40 escape/avoidance trials response to inescapable shock in the har-

in the shuttle box. ness. This contrasts with the interfer-

Figures 1, 2, and 3 present the ence effect produced by inescapable shock

in 5s which have had no prior experience

escape/avoidance performance of the with shock or in 5s which have had prior

Preinescapable group on Day 3. In experience with inescapable shock. Thus,

general this group performed like the if an animal first learns that its re-

8 MARTIN E. P. SELIGMAN AND STEVEN F. MAIER



spending produces shock termination and cause 5 learned that shock termination

then faces a situation in which reinforce- was independent of its responses.

ment is independent of its responding, it Learning that one's own responding

is more persistent in its attempts to es- and reinforcement are independent might

cape shock than is a naive animal. be expected to play a role in appetitive

situations. If S received extensive pre-

GENERAL DISCUSSION training with rewarding brain stimula-

tion delivered independently of its operant

We have proposed that 5" learned as a responding, would the subsequent acquisi-

consequence of inescapable shock that its tion of a bar press to obtain this reward

responding was independent of shock be retarded? Further, might learned

termination, and therefore the probability "helplessness" transfer from aversive to

of response initiation during shock de- appetitive situations or vice versa?

creased. Alternative explanations might

If dogs learn in one situation that

be offered: (a) Inactivity, somehow, re-

their active responding is to no avail,

duces the aversiveness of shock. Thus 5

and then transfer this training to another

failed to escape shock in the shuttle box

shock situation, the opposite type of

because it had been reinforced for inac-

transfer (avoidance learning sets) might

tivity in the harness. Since the inter-

be possible: If a dog first learned a bar-

ference effect occurred in 5s which had

rier-hurdling response which avoided

been curarized during inescapable shock,

shock in the shuttle box, would that dog

such an aversiveness-reducing mechanism

be facilitated in learning to panel press

would have to be located inward of the

to avoid shock in the harness (to a dif-

neuro-myal junction. (&) S failed to es-

cape in the shuttle box because certain ferent CS) ? Our finding, that dogs

responses which facilitate barrier jumping which first successfully escape shock in

were extinguished in the harness during the shuttle box later showed enhanced

panel pressing in the harness, is conso-

inescapable shock. In conventional ex-

tinction procedures, some response is first nant with this prediction.

explicitly reinforced by correlation with Does learning about response—rein-

shock termination, and then that response forcement contingencies have its analogs

is extinguished by removing shock alto- in classical conditioning? If 6" experi-

gether from the situation. Responding enced two stimuli randomly interspersed

during extinction is conventionally not with each other (adventitious pairings

Mwcorrelated with shock termination; possible), would it be retarded in form-

rather, responding is correlated with the ing an association between the two stim-

total absence of shock. In our harness uli once true pairing was begun? Con-

situation, no response was first explicitly versely, pretraining in which one stimulus

reinforced, and shock was presented is correlated with a US might facilitate

throughout the session. A broader con- the acquisition of the CR to a new CS.

cept of extinction, however, might be Pavlov (1927, p. 75) remarked that the

tenable. On this view, any procedure first establishment of a conditioned in-

which decreases the probability of a re- hibitor took longer than any succeeding

sponse by eliminating the incentive to one.

respond is an extinction procedure. If In conclusion, learning theory has

the independence of shock termination stressed that two operations, explicit

and responding eliminates the incentive contiguity between events (acquisition)

to respond (as assumed), then our har- and explicit noncontiguity (extinction),

ness procedure could be thought of as an produce learning. A third operation that

extinction procedure. Such an explana- is proposed, independence between events,

tion seems only semantically different also produces learning, and such learning

from the one we have advanced, since may have effects upon behavior that dif-

both entail that the probability of re- fer from the effects of explicit pairing

sponding during shock has decreased be- and explicit nonpairing. Such learning

ESCAPE FAILURE WITH TRAUMATIC SHOCK



may produce an 5 who does not attempt MOWRER, O. H. Learning theory and be-

to escape electric shock; an 51 who, even havior. New York: Wiley, 1960.

if he does respond, may not benefit from OVERMIER, J. B., & SELIGMAN, M. E. P. Ef-

instrumental contingencies. fects of inescapable shock on subsequent

escape and avoidance learning. J. comp.

REFERENCES physiol. Psychol., 1967, 63, 28-33.

BRUSH, F. R., MYER, J. S., & PALMER, M. E. PAVLOV, I. P. Conditioned reflexes. New

Effects of kind of prior training and inter- York: Dover, 1927.

session interval upon subsequent avoidance RICHTER, C. On the phenomenon of sudden

learning. /. comp. physiol. Psychol., 1963, death in animals and man. Psychosom.

56, 539-S4S. Med., 19S7, 19, 191-198.

COFER, C. N., & APPLEY, M. H. Motivation: SOLOMON, R. L., & WYNNE, L. C. Trau-

Theory and research. New York: Wiley, matic avoidance learning: Acquisition in

1964. normal dogs. Psychol. Monogr., 1953,

LEFCOURT, H. M. Internal vs. external con- 67(4, Whole No. 3S4).

trol of reinforcement: A review. Psychol.

Bull, 1966, 65, 206-221. (Received June 1, 1966)