BT. J . P~ychol.(1976), 67,4, p p . 529-635. Printed in Ureat B d i n
529
THE EFFECT OF EMBEDDED WORDS I N A BRIEF VISUAL DISPLAY BY DAVID E . SOMEKH Formerly: Department of Psychology, Bedford College* In a replication and extension of an experiment by Eagle, Wolitzky & Klein (1966), subjects wrote brief stories describing en Object Relations Test card following exposure to a 7 x 7 letter matrix in which were embedded either neutral words or emotive words. Of six groups of subjects who participated, two groups were exposed to the matrioes for 10 sec (‘supraliminal control condition’), and the remaining four groups for 1 BBO (‘subliminal condition’). Independent judges were able to distinguish, to a significent degree, between the stories of subjects who were exposed to emotive words and those exposed to n13Utr8l words under the ‘8ubliminal condition’; they could not distinguish, however, between the stories of subjects exposed for the longer duration (‘suprdiminsl condition’) to emotive and neutral worde respectively. A crude content snalysie yielded signilicantly more unpleeeant words in stories from the eubgroups exposed to emotive rather than neutral worda, but again only under the very brief exposure condition, the difference being absent in supraliminel control groups. The relevance of these f h d i n p to previous experimmta is discussed.
There is evidence to show that stimuli which may be unperceived at the time of their occurrence can be retrieved under conditions which favour the operation of primary thought processes (Fisher, 1960; Silverman & Silverman, 1964; Dixon, 1971).
Eagle et al. (1966) looked specifically at the effects of a part of the stimulus display of which the observer was not aware. They used an embedded figure (a duck in a tree) which produced significantly more ‘duck related’ items in subsequent imagining of a nature scene than did a control figure, even though subjects were unaware of the source of their imagery. They concluded ‘. . . connotative aspects of the ground, although apparently unperceived, may become manifest in non-perceptual response modes such as imagery’ (p. 839). Of importance t o selective attention models is the level of analysis which such an effect implies. It was decided to use a similar technique to Eagle et al. but to take their study a stage further in two ways. First, a within awareness control group was used to attempt to show a qualitative difference between the two situations and second, embedded emotive word8 were used, with words of very similar structure but different meaning as controls. METHOD Subjects were tested in groups, subjects within each group being arbitrarily divided into two equal subgroups labelled ‘A‘ and ‘B’ respectively. Group A (experimental group) were briefly exposed t o a matrix of letters with two emotive words, wcxand PAIN,embedded in it. Group B (control group) were shown a very similar matrix except t h a t the embedded words were DUOK and RAIN. Both groups then wrote down BB much of the matrices as they could remember and then t h e Grst ten words t h a t came into their heads. The whole group was then shown an Object Relations Test card and asked to write a brief story about it. Stories were collected, typed and randomized and blind judges with knowledge of the procedure were asked t o assign * Now a t The Maudsley Hospital, Denmark Hill, London, S.E.6.
D. E. SOMEKH
530 R T Z W J Q L
X H G B E*Y J Q ,3 V P A D B M T H R Y G T X F U M Z D J
R L I
D R N
Q V C B
Z
S K M
P Y Z W J Q L
X
H
O*T S V F B H P T X M Z
G B P F J Q L 11 R A I N M T Q Z Y G V S D U C K F J B M
Fig. 1. Stimuli used (white upper case letters on blmk background except for two red letter8 marked thus *).
the stories, on the 1)asisof their content, to group A or group B. It waa argued that the emotive words would a f f e d the content of the stories given by group A sufficiently for the judges to be able to distinguish between stories from the two subgroups to a signiscant extent. I n a pilot experiment a 6 x 6 letter matrix, with a red letter in one of the four quadrants and a three-letter neutral word in differing relation to it, had been shown to 40 subjects tachistoscopically. Subjects pressed a thumb-switch which illuminated a card bearing the matrix and started a stop-clok The subject then reported the initial target (red letter) verbally and pressed the thumb -switchwhen he or she found the secondary target (aneutral word - reporting it verbally as a check), thus stopping the clock to register total search time. The fasted time from 240 trials (12 x 20 subjects), when the word and letter were not actually adjacent, was 1.38 sec (mean time 2.91 sec). It was therefore argued that With a 7 x 7 letter matrix and the subject not aware that words were present in the matrix, a 1 sec exposure, even if repeated, would make it vii.tually impossible for subjects to report the embedded words. On the other hand in view of T‘arbus’s finding (1967)that during a 6 sec period the observer of a complex picture changed points of b a t i o n 18 times, it is likely under these conditions that subjects would have perfo~medseveral aaccadea during the 1 sec period and therefore an image of the area of the matria. containing the embedded words would have fallen on the retina.
M&W The stimuli arc, shown in Fig. 1. These and Object Relations Test cards B 2 and B3 were mounted as 2 x 2 in. slides and projected on t o a screen at an average distance of 12 f t 6 in. from the projector. A 300 watt Leieegang magazine-loadingprojector waa used with a Prontor shutter connected via a voltage stabilizer to a Devices digither. Subjects were given protocol sheets which mendy had a 7 x 7 open matrix in the top left corner and below, the numbers 1-10, for the association exercise. Subjects
One hundred e n d twenty-four subjects (42 male, 82 female), mainly undergraduates (some of them in psychdogy) and sixth-formers. Procedure Subjects were issted in groups of between 16 and 24. Alternate subjects were handed protocol sheets and told ‘ you belong to group A (or B), please write your letter in the right-hand corner of the sheet’. Having divided subjects into equal groups and checked that no subject was red colour-blind, the experimenter said, ‘I am now going to show you a matrix of letters. Somewhere in the matrix will be a red letter which is your primary target. As soon aa you have found the red letter, try and see aa much of the reet of the matrix aa you can. 1’11 show you the matrix very briefly twice, separated by a few seconds. Immediately afferwards I want you to write dcsm as much of the matrix as you can remember in the s p e provided. Write aa many letters as you can, even if you’re not quife sure they’re right, beginning with the red letter. Now, hem’s the diBicult bit, for which I need your Cooperation. I’m going to show the matrix to group A first and while I’m doing this I want p u p B to look down at the desk or anyway, promise not to look. As soon aa group A are trying to write down what they mw, I shell show group B a similar matrix. So the procedure will be: group A look a t the board, group B look dclwn. “hen, group A try and write down what they aaw while group B look a t the board. Any questions?’
Embedded words in a brief visual display
531
At the command, ‘group A look a t the board, group B look down’, group A were shown one of the two matrices twice for 1.0 sec, separated by a gap of 3 sec. Approximately 15 sec after group A were first shown their matrix, group B were shown the other of the two matrices. After approximately 2 min subjects were told, ‘If you’ve written down &B much ea you can remember, now write down the first ten words which come into your head, in the space provided. When you have finished, turn your sheet over.’ When all subjects had finished, they were told, ‘I am now going to show you a picture. After you have seen it I want you to write a short s t o r y , about 6-8 lines, about what you think is happening - that is, I’d like you to write about two lines on how you think the situation came about, four to six lines on what is happening now and about two lines on what you think will happen next’. Subjects were shown one of the two ORT pictures for 10 sec and then wrote their stories. Protocols were collected and subjects were asked what they thought the experiment was about and its true nature was explained. Four groups of subjects, totalling 80 subjects, 40 experimental and 40 control, were tested in this manner. Order of presentation of neutral and emotive matrices and use of ORT card B2 or B3 were randomized across groups. On checking the protocols, without exception all subjects had written down the correct red letter (either an E or an 0)correspondingto their status aa group A or B. No subject had written down either of the embedded words. On questioning. no subject was aware that embedded words had been present. I n the ‘supraliminal’ session, two groups of subjects, numbering 20 and 24, were tested. The procedure waa exactly the same except that each exposure of the matrices was for 10 sec instead of 1 sec. By questioning and from scrutiny of the protocols it was established that a total of ten subjects of the 44 had seen and reported one or both of the stimulus words. PAIN waa reported five times, FUCK four times, RAIN three times and DUCK not a t all.
RESULTS
The stories for each group were randomized and typed and were sent to five judges with details of procedure. The judges’ task was to suggest, by reference to story content, which came from the experimental subgroup and which from the control subgroup. The five judges for each group of stories were two naive judges (NJ), usually psychology postgraduates with no detailed knowledge of the Object Relations Test technique; two expert judges (EJ),psychoanalysts with knowledge of the ORT, and one semi-experienced judge (SJ),usually a psychology lecturer with some knowledge of the ORT. A total of 14 judges were used for the 30 judgements (five judges for each of the six experimental groups). Table 1 shows the number of correct judgements by each of the five judges for each group. To test for significance, the results of each group of stories were organized into a contingency table according t o the number of judges making a correct judgement on a particular story from group A or group B. Each table was then analysed for significance, using Kendall’s tau (with correction for ties) and the result expressed as a z score. A cumulative z score over successive groups was then obtained. Over a total of 80 subjects in the ‘subliminal condition’, judges were able to distinguish between the two groups to a significant extent, z, = 2-027 (one-tailed, P < 0.025). When matrices were shown to subjects for a longer period, judges were unable to distinguish between stories coming from the 44 subjects, z, = 0.071 (n.s.). These findings lend tentative support to the suggestion that stories given after brief exposure to hidden emotive words are different from those following exposure to hidden neutral words and that longer exposure to the words reduces the effect. The exact nature of the difference is difficult to determine as each judge used his own criterion for assessment. There was, nevertheless, general agreement that stories
D. E. SOMEKH Table 1. Summary of story judgements (Numbero o m t l y assigned to group A or B)
‘Subliminal’ d o n e Uroup 1 Qroup 2 Uroup 3 Uroup 4 ‘Supralimins1’ eessiozls Uroup 6 Group 6
Number of eubjecta , h A + B NJ1
NJI
EJ,
EJa
SJ
Kendall’e 8
2
2,
8 12 8 10
10 16 10 14
10 16 10 12
10 16 10 16
8 12 10 12
19 47 23 16
1.003 1.370 1.217 0.464
1.003 1.680 2.073 2.027
16 24 16 24
Judges 7
Cumulative
(one-tailed, P < 0.026) 20 24
10 14
10 12
8 14
8 12
14 10
-2 6
-0.076 -0.076 0.176
0.071 (n.8.)
following brief exposure to the words FUCE and P ~ I Nshowed more disturbance, whether violent themes, uncertainty, conflict or loss, than stories following exposure to the wor&.DUOKand RAIN.I n an attempt to provide a more objective demonstration of differcmces between the two groups of stories, a crude content analysis of the 124 stones was performed. Content analysis
From 124 stories, average length 80 words, approximately 1600 different words were extracted. These were randomized and presented to seven independent judges who were asked to rate them as pleasant, neutral or unpleasant. It was decided to take as a criterion a minimum of three out of seven judgements of U or P,with U and P judgemerits cancelling each other out if they co-occurred. In fact, of the 349 words qualifying as U or P words, only 17 (< 5 per cent) were not uniformly judged by a minimum of three judges (the others judging the word as neutral). The stories were then scored for content of unpleasant or pleasant words. Overall, story lengths and proportion of non-neutral words were similar across groups. However, in the ‘subliminal’ sessions, stories from the experimental group A tended to have more unpleasant than pleasant words, the reverse being true for stories from control group B. The only exception was in group 2 where stories from group A also had more pleasmt than unpleasant words. Overall, in experimental condition stories 45 per cent (112) of non-neutral words were pleasant, as compared to 60 per cent (153)from control condition stories. In the ‘supraliminal’ sessions, the proportion of pleasant to unpleasant words did not differ for stories from group A and B, the proportion of pleasmt words being 56 per cent (58) and 52 per cent (71) respectively. I n order to assess the significance of these findings, following a suggestion from Mr J. D. Valentine of Bedford College, each story was assigned a ‘difference score’ which represenved the proportion of pleasant to unpleasant words, as follows:
and could therefore vary between
- 1 and + 1.
Embedded
woTdp
in a brief visual display
633
Table 2 . Content analysis: man pleasant/u.rajpleaaant index scores (S.D.B in parenthesea)
Group A (experimental)
‘Subliminal’ sessions Group 1 Group 2 Uroup 3 Group 4
-0.20 +0.16 -0.66 -0.18
(0.77) (0.68) (0.67) (0.71)
Group B (control)
(4
U
P
8x2 12 x 2 8x 2 12 ~2
20 70 11 40
0.12 0.56 0.01 0.04
-0.04 (0.13) 10 x 2 +0*03 (0.80) 12 X 2
66
0.66 0-16
+0.31 (0.81) +0.12 (0.66) +043 (0.72) +0*30 (0-76)
‘Supraliminal ’ seesions Group 6 Group 6
+0*06 (0.30) +0*06 (0.68)
66
--
Table 3. Frequency of U and P words expressed a8 a percentage of the total number of words in the sibrim Experimental condition
‘Subliminal’ sessions ‘Supraliminal’ sessions
Control condition
%U
%P
%U
%P
4.24 2.86
3.41 3.36
3.30 3.60
5.00 3.82
Within each of the sessions indices from experimental and control stories were compared, using the Mann-Whitney U test. Table 2 shows the mean index values for each group, the U values comparing indices within each session and associated one-tailed probabilities. Probabilities were summed across sessions, using the formula xg = - 2 Zflog, Pi, which yielded, for the ‘subliminalysessions,
X 2(d.f. 8) = 20-48 (P < 0.01) and for the ‘supraliminal’ sessions 4.48 (P> 0.3) Table 3 shows subliminal and supraliminal session totals expressed as a percentage of the total number of words in the stories. It appears from the table that there is a tendency under the subliminal condition for emotive stimuli such as FTJCK and PAIN to produce stories with more unpleasant words, while neutral stimuli such as DUCK and RAIN produce stories with more pleasant words. When the words approach awareness, i.e. some subjects are able to report them, the percentage of unpleasant words in the experimental condition and of pleasant words in the control condition both decreased; the percentage of pleasant words in the experimental condition and of unpleasant words in the control condition remained unchanged, however. The larger fall in the percentage of unpleasant words in experimental (group A) supraliminal stories might tentatively be ascribed to a defensive response, suggesting a quite different handling of the emotive stimuli when they are on the borders of awareness than when subjects are totally unaware of them.
X2(d.f. 4)
=
634
D. E. SOMEKH DISCUSSION
The findings appear to confirm those of many previous workers (see reviews by Fisher, 1960 and. Dixon, 1971) that marginal or incidental stimuli, like subliminal stimuli, may be registered without awareness and can be recovered in a variety of modes of experience and behaviour - in this case, symbolically transferred as imagery in a story. The use of suitable controls makes explanations of the phenomenon in terms of partial cues of ‘fleeting’ perceptions improbable. As Eagle et d.comment, perceptions too fleeting to be reported would seem to be operationally indistinguishable from stimulus registrations without awareness. It can be argued that some of the subjects in the ‘supraliminal’ sessions should have been responding to the stimuli in the same way as subjects in the ‘subliminal’ sessions, as they were unable to report the stimuli subsequently. The z scores for judgements of groups 5 and 6 are, however, both lower than that for any of the ‘subliminal’ groups, and it would appear that the incidental stimuli had a much weaker effect when presented near the awareness threshold. This is consistent with the findings of Paul & Fisher (1969) and others, that the subliminal effect increases as stimuli are presented further below threshold. In the present, experiment a considerable number of factors contribute to error variance. To give two examples, the pleasankunpleasant measure is very rough and ready; and secondly, the word RAIN may have a strong effect with some subjects when followed by OR’C card B2 which shows shadowy figures apparently sheltering under a tree outside a large house (Phillipson, 1973). Rather than detract from the findings, this observation suggests that the possibility for a more sophisticated methodology which may contribute better to an understanding of the mechanism underlying the effect and facilitate a considerable reduction in error variance. For example, if the individual subjects were tested with the ORT before the experiment, to obtain baseline profiles, the specific effect of the incidental stimuli could then be assessed and this procedure would enable use of objective scoring methods, which are available (Phillipson, 1973). The implications for theories of selective attention are twofold. First, consistent with the results of previous studies the findings tend to support the view that incidental stimuli may be recovered or have an effect after a period exceeding that normally associated with short-term memory. Second, the difference between experimental (group A) and control (group B) conditions appears to depend on subjects discriminating words of very similar structure but different meaning. This supports Gndings from previous experiments (e.g. Somekh & Wilding, 1973), which suggest that incidental or ‘iiielevant’ stimuli are processed to a high level of analysis. This experiment was supported by a grant from the Social Science Research Council as part of the work for doctorial thesis (London University). The author would like to thank Dr J. Sandler for helpful suggestions and Dr J.M. Wilding for encourrtgement and advice. The author would ale3 like to thank the postgraduates who acted as judgm and (in alphabetical order) Mm P. Antonis, Mr V. B. Kentor, Dr M. Lawlor, Mr H. Phillipson end Mrs A. M. Sandler. Special thanka to Mrs S. Fien.
Embedded words in a brief visual display
535
REFERENCES DIXON,N. F. (1971). Sublimiraal Perceptirm. London: McGraw-Hill. EAGLE, M., WOLITZKY, D. I. & WIN, G. S. (1966). Imagery: effect of a concealed figure in B stimulus. Science, N . Y . 151,837-839. FISHER, C. (1960). Preconscious stimulation in dream, associations and images. Itlonogr. 7: Psychol. I s m s 2, 1-40. PAUL,I. H. & FISHER,c. (1959). S u b l d a l visual stimulation: a study of its influence on subsequent images and dream. J . new. me&. Db. 129, 316-340. H. (1973). Person81 communication. PHILLIPSON, SILVERWAN, L. H. t SILVERMAN, D. K. (1964). A clinical-experimental approach to the study of subliminal stimulation. J . abraorm. SOC.Psychol. 69, 158-172. Soram, D. E. & WILDING, J. M. (1973). Perception without awareness in a dichoptic viewing situation. Br. J . Psychol. 64, 339-349. YARBUS, A. L. (1967). Eye-movements and Viaion. New York: Plenum Preess.
(Manwcript received 29 July 1974; revised mnuscri.pt received 16 September 1975)
529
THE EFFECT OF EMBEDDED WORDS I N A BRIEF VISUAL DISPLAY BY DAVID E . SOMEKH Formerly: Department of Psychology, Bedford College* In a replication and extension of an experiment by Eagle, Wolitzky & Klein (1966), subjects wrote brief stories describing en Object Relations Test card following exposure to a 7 x 7 letter matrix in which were embedded either neutral words or emotive words. Of six groups of subjects who participated, two groups were exposed to the matrioes for 10 sec (‘supraliminal control condition’), and the remaining four groups for 1 BBO (‘subliminal condition’). Independent judges were able to distinguish, to a significent degree, between the stories of subjects who were exposed to emotive words and those exposed to n13Utr8l words under the ‘8ubliminal condition’; they could not distinguish, however, between the stories of subjects exposed for the longer duration (‘suprdiminsl condition’) to emotive and neutral worde respectively. A crude content snalysie yielded signilicantly more unpleeeant words in stories from the eubgroups exposed to emotive rather than neutral worda, but again only under the very brief exposure condition, the difference being absent in supraliminel control groups. The relevance of these f h d i n p to previous experimmta is discussed.
There is evidence to show that stimuli which may be unperceived at the time of their occurrence can be retrieved under conditions which favour the operation of primary thought processes (Fisher, 1960; Silverman & Silverman, 1964; Dixon, 1971).
Eagle et al. (1966) looked specifically at the effects of a part of the stimulus display of which the observer was not aware. They used an embedded figure (a duck in a tree) which produced significantly more ‘duck related’ items in subsequent imagining of a nature scene than did a control figure, even though subjects were unaware of the source of their imagery. They concluded ‘. . . connotative aspects of the ground, although apparently unperceived, may become manifest in non-perceptual response modes such as imagery’ (p. 839). Of importance t o selective attention models is the level of analysis which such an effect implies. It was decided to use a similar technique to Eagle et al. but to take their study a stage further in two ways. First, a within awareness control group was used to attempt to show a qualitative difference between the two situations and second, embedded emotive word8 were used, with words of very similar structure but different meaning as controls. METHOD Subjects were tested in groups, subjects within each group being arbitrarily divided into two equal subgroups labelled ‘A‘ and ‘B’ respectively. Group A (experimental group) were briefly exposed t o a matrix of letters with two emotive words, wcxand PAIN,embedded in it. Group B (control group) were shown a very similar matrix except t h a t the embedded words were DUOK and RAIN. Both groups then wrote down BB much of the matrices as they could remember and then t h e Grst ten words t h a t came into their heads. The whole group was then shown an Object Relations Test card and asked to write a brief story about it. Stories were collected, typed and randomized and blind judges with knowledge of the procedure were asked t o assign * Now a t The Maudsley Hospital, Denmark Hill, London, S.E.6.
D. E. SOMEKH
530 R T Z W J Q L
X H G B E*Y J Q ,3 V P A D B M T H R Y G T X F U M Z D J
R L I
D R N
Q V C B
Z
S K M
P Y Z W J Q L
X
H
O*T S V F B H P T X M Z
G B P F J Q L 11 R A I N M T Q Z Y G V S D U C K F J B M
Fig. 1. Stimuli used (white upper case letters on blmk background except for two red letter8 marked thus *).
the stories, on the 1)asisof their content, to group A or group B. It waa argued that the emotive words would a f f e d the content of the stories given by group A sufficiently for the judges to be able to distinguish between stories from the two subgroups to a signiscant extent. I n a pilot experiment a 6 x 6 letter matrix, with a red letter in one of the four quadrants and a three-letter neutral word in differing relation to it, had been shown to 40 subjects tachistoscopically. Subjects pressed a thumb-switch which illuminated a card bearing the matrix and started a stop-clok The subject then reported the initial target (red letter) verbally and pressed the thumb -switchwhen he or she found the secondary target (aneutral word - reporting it verbally as a check), thus stopping the clock to register total search time. The fasted time from 240 trials (12 x 20 subjects), when the word and letter were not actually adjacent, was 1.38 sec (mean time 2.91 sec). It was therefore argued that With a 7 x 7 letter matrix and the subject not aware that words were present in the matrix, a 1 sec exposure, even if repeated, would make it vii.tually impossible for subjects to report the embedded words. On the other hand in view of T‘arbus’s finding (1967)that during a 6 sec period the observer of a complex picture changed points of b a t i o n 18 times, it is likely under these conditions that subjects would have perfo~medseveral aaccadea during the 1 sec period and therefore an image of the area of the matria. containing the embedded words would have fallen on the retina.
M&W The stimuli arc, shown in Fig. 1. These and Object Relations Test cards B 2 and B3 were mounted as 2 x 2 in. slides and projected on t o a screen at an average distance of 12 f t 6 in. from the projector. A 300 watt Leieegang magazine-loadingprojector waa used with a Prontor shutter connected via a voltage stabilizer to a Devices digither. Subjects were given protocol sheets which mendy had a 7 x 7 open matrix in the top left corner and below, the numbers 1-10, for the association exercise. Subjects
One hundred e n d twenty-four subjects (42 male, 82 female), mainly undergraduates (some of them in psychdogy) and sixth-formers. Procedure Subjects were issted in groups of between 16 and 24. Alternate subjects were handed protocol sheets and told ‘ you belong to group A (or B), please write your letter in the right-hand corner of the sheet’. Having divided subjects into equal groups and checked that no subject was red colour-blind, the experimenter said, ‘I am now going to show you a matrix of letters. Somewhere in the matrix will be a red letter which is your primary target. As soon aa you have found the red letter, try and see aa much of the reet of the matrix aa you can. 1’11 show you the matrix very briefly twice, separated by a few seconds. Immediately afferwards I want you to write dcsm as much of the matrix as you can remember in the s p e provided. Write aa many letters as you can, even if you’re not quife sure they’re right, beginning with the red letter. Now, hem’s the diBicult bit, for which I need your Cooperation. I’m going to show the matrix to group A first and while I’m doing this I want p u p B to look down at the desk or anyway, promise not to look. As soon aa group A are trying to write down what they mw, I shell show group B a similar matrix. So the procedure will be: group A look a t the board, group B look dclwn. “hen, group A try and write down what they aaw while group B look a t the board. Any questions?’
Embedded words in a brief visual display
531
At the command, ‘group A look a t the board, group B look down’, group A were shown one of the two matrices twice for 1.0 sec, separated by a gap of 3 sec. Approximately 15 sec after group A were first shown their matrix, group B were shown the other of the two matrices. After approximately 2 min subjects were told, ‘If you’ve written down &B much ea you can remember, now write down the first ten words which come into your head, in the space provided. When you have finished, turn your sheet over.’ When all subjects had finished, they were told, ‘I am now going to show you a picture. After you have seen it I want you to write a short s t o r y , about 6-8 lines, about what you think is happening - that is, I’d like you to write about two lines on how you think the situation came about, four to six lines on what is happening now and about two lines on what you think will happen next’. Subjects were shown one of the two ORT pictures for 10 sec and then wrote their stories. Protocols were collected and subjects were asked what they thought the experiment was about and its true nature was explained. Four groups of subjects, totalling 80 subjects, 40 experimental and 40 control, were tested in this manner. Order of presentation of neutral and emotive matrices and use of ORT card B2 or B3 were randomized across groups. On checking the protocols, without exception all subjects had written down the correct red letter (either an E or an 0)correspondingto their status aa group A or B. No subject had written down either of the embedded words. On questioning. no subject was aware that embedded words had been present. I n the ‘supraliminal’ session, two groups of subjects, numbering 20 and 24, were tested. The procedure waa exactly the same except that each exposure of the matrices was for 10 sec instead of 1 sec. By questioning and from scrutiny of the protocols it was established that a total of ten subjects of the 44 had seen and reported one or both of the stimulus words. PAIN waa reported five times, FUCK four times, RAIN three times and DUCK not a t all.
RESULTS
The stories for each group were randomized and typed and were sent to five judges with details of procedure. The judges’ task was to suggest, by reference to story content, which came from the experimental subgroup and which from the control subgroup. The five judges for each group of stories were two naive judges (NJ), usually psychology postgraduates with no detailed knowledge of the Object Relations Test technique; two expert judges (EJ),psychoanalysts with knowledge of the ORT, and one semi-experienced judge (SJ),usually a psychology lecturer with some knowledge of the ORT. A total of 14 judges were used for the 30 judgements (five judges for each of the six experimental groups). Table 1 shows the number of correct judgements by each of the five judges for each group. To test for significance, the results of each group of stories were organized into a contingency table according t o the number of judges making a correct judgement on a particular story from group A or group B. Each table was then analysed for significance, using Kendall’s tau (with correction for ties) and the result expressed as a z score. A cumulative z score over successive groups was then obtained. Over a total of 80 subjects in the ‘subliminal condition’, judges were able to distinguish between the two groups to a significant extent, z, = 2-027 (one-tailed, P < 0.025). When matrices were shown to subjects for a longer period, judges were unable to distinguish between stories coming from the 44 subjects, z, = 0.071 (n.s.). These findings lend tentative support to the suggestion that stories given after brief exposure to hidden emotive words are different from those following exposure to hidden neutral words and that longer exposure to the words reduces the effect. The exact nature of the difference is difficult to determine as each judge used his own criterion for assessment. There was, nevertheless, general agreement that stories
D. E. SOMEKH Table 1. Summary of story judgements (Numbero o m t l y assigned to group A or B)
‘Subliminal’ d o n e Uroup 1 Qroup 2 Uroup 3 Uroup 4 ‘Supralimins1’ eessiozls Uroup 6 Group 6
Number of eubjecta , h A + B NJ1
NJI
EJ,
EJa
SJ
Kendall’e 8
2
2,
8 12 8 10
10 16 10 14
10 16 10 12
10 16 10 16
8 12 10 12
19 47 23 16
1.003 1.370 1.217 0.464
1.003 1.680 2.073 2.027
16 24 16 24
Judges 7
Cumulative
(one-tailed, P < 0.026) 20 24
10 14
10 12
8 14
8 12
14 10
-2 6
-0.076 -0.076 0.176
0.071 (n.8.)
following brief exposure to the words FUCE and P ~ I Nshowed more disturbance, whether violent themes, uncertainty, conflict or loss, than stories following exposure to the wor&.DUOKand RAIN.I n an attempt to provide a more objective demonstration of differcmces between the two groups of stories, a crude content analysis of the 124 stones was performed. Content analysis
From 124 stories, average length 80 words, approximately 1600 different words were extracted. These were randomized and presented to seven independent judges who were asked to rate them as pleasant, neutral or unpleasant. It was decided to take as a criterion a minimum of three out of seven judgements of U or P,with U and P judgemerits cancelling each other out if they co-occurred. In fact, of the 349 words qualifying as U or P words, only 17 (< 5 per cent) were not uniformly judged by a minimum of three judges (the others judging the word as neutral). The stories were then scored for content of unpleasant or pleasant words. Overall, story lengths and proportion of non-neutral words were similar across groups. However, in the ‘subliminal’ sessions, stories from the experimental group A tended to have more unpleasant than pleasant words, the reverse being true for stories from control group B. The only exception was in group 2 where stories from group A also had more pleasmt than unpleasant words. Overall, in experimental condition stories 45 per cent (112) of non-neutral words were pleasant, as compared to 60 per cent (153)from control condition stories. In the ‘supraliminal’ sessions, the proportion of pleasant to unpleasant words did not differ for stories from group A and B, the proportion of pleasmt words being 56 per cent (58) and 52 per cent (71) respectively. I n order to assess the significance of these findings, following a suggestion from Mr J. D. Valentine of Bedford College, each story was assigned a ‘difference score’ which represenved the proportion of pleasant to unpleasant words, as follows:
and could therefore vary between
- 1 and + 1.
Embedded
woTdp
in a brief visual display
633
Table 2 . Content analysis: man pleasant/u.rajpleaaant index scores (S.D.B in parenthesea)
Group A (experimental)
‘Subliminal’ sessions Group 1 Group 2 Uroup 3 Group 4
-0.20 +0.16 -0.66 -0.18
(0.77) (0.68) (0.67) (0.71)
Group B (control)
(4
U
P
8x2 12 x 2 8x 2 12 ~2
20 70 11 40
0.12 0.56 0.01 0.04
-0.04 (0.13) 10 x 2 +0*03 (0.80) 12 X 2
66
0.66 0-16
+0.31 (0.81) +0.12 (0.66) +043 (0.72) +0*30 (0-76)
‘Supraliminal ’ seesions Group 6 Group 6
+0*06 (0.30) +0*06 (0.68)
66
--
Table 3. Frequency of U and P words expressed a8 a percentage of the total number of words in the sibrim Experimental condition
‘Subliminal’ sessions ‘Supraliminal’ sessions
Control condition
%U
%P
%U
%P
4.24 2.86
3.41 3.36
3.30 3.60
5.00 3.82
Within each of the sessions indices from experimental and control stories were compared, using the Mann-Whitney U test. Table 2 shows the mean index values for each group, the U values comparing indices within each session and associated one-tailed probabilities. Probabilities were summed across sessions, using the formula xg = - 2 Zflog, Pi, which yielded, for the ‘subliminalysessions,
X 2(d.f. 8) = 20-48 (P < 0.01) and for the ‘supraliminal’ sessions 4.48 (P> 0.3) Table 3 shows subliminal and supraliminal session totals expressed as a percentage of the total number of words in the stories. It appears from the table that there is a tendency under the subliminal condition for emotive stimuli such as FTJCK and PAIN to produce stories with more unpleasant words, while neutral stimuli such as DUCK and RAIN produce stories with more pleasant words. When the words approach awareness, i.e. some subjects are able to report them, the percentage of unpleasant words in the experimental condition and of pleasant words in the control condition both decreased; the percentage of pleasant words in the experimental condition and of unpleasant words in the control condition remained unchanged, however. The larger fall in the percentage of unpleasant words in experimental (group A) supraliminal stories might tentatively be ascribed to a defensive response, suggesting a quite different handling of the emotive stimuli when they are on the borders of awareness than when subjects are totally unaware of them.
X2(d.f. 4)
=
634
D. E. SOMEKH DISCUSSION
The findings appear to confirm those of many previous workers (see reviews by Fisher, 1960 and. Dixon, 1971) that marginal or incidental stimuli, like subliminal stimuli, may be registered without awareness and can be recovered in a variety of modes of experience and behaviour - in this case, symbolically transferred as imagery in a story. The use of suitable controls makes explanations of the phenomenon in terms of partial cues of ‘fleeting’ perceptions improbable. As Eagle et d.comment, perceptions too fleeting to be reported would seem to be operationally indistinguishable from stimulus registrations without awareness. It can be argued that some of the subjects in the ‘supraliminal’ sessions should have been responding to the stimuli in the same way as subjects in the ‘subliminal’ sessions, as they were unable to report the stimuli subsequently. The z scores for judgements of groups 5 and 6 are, however, both lower than that for any of the ‘subliminal’ groups, and it would appear that the incidental stimuli had a much weaker effect when presented near the awareness threshold. This is consistent with the findings of Paul & Fisher (1969) and others, that the subliminal effect increases as stimuli are presented further below threshold. In the present, experiment a considerable number of factors contribute to error variance. To give two examples, the pleasankunpleasant measure is very rough and ready; and secondly, the word RAIN may have a strong effect with some subjects when followed by OR’C card B2 which shows shadowy figures apparently sheltering under a tree outside a large house (Phillipson, 1973). Rather than detract from the findings, this observation suggests that the possibility for a more sophisticated methodology which may contribute better to an understanding of the mechanism underlying the effect and facilitate a considerable reduction in error variance. For example, if the individual subjects were tested with the ORT before the experiment, to obtain baseline profiles, the specific effect of the incidental stimuli could then be assessed and this procedure would enable use of objective scoring methods, which are available (Phillipson, 1973). The implications for theories of selective attention are twofold. First, consistent with the results of previous studies the findings tend to support the view that incidental stimuli may be recovered or have an effect after a period exceeding that normally associated with short-term memory. Second, the difference between experimental (group A) and control (group B) conditions appears to depend on subjects discriminating words of very similar structure but different meaning. This supports Gndings from previous experiments (e.g. Somekh & Wilding, 1973), which suggest that incidental or ‘iiielevant’ stimuli are processed to a high level of analysis. This experiment was supported by a grant from the Social Science Research Council as part of the work for doctorial thesis (London University). The author would like to thank Dr J. Sandler for helpful suggestions and Dr J.M. Wilding for encourrtgement and advice. The author would ale3 like to thank the postgraduates who acted as judgm and (in alphabetical order) Mm P. Antonis, Mr V. B. Kentor, Dr M. Lawlor, Mr H. Phillipson end Mrs A. M. Sandler. Special thanka to Mrs S. Fien.
Embedded words in a brief visual display
535
REFERENCES DIXON,N. F. (1971). Sublimiraal Perceptirm. London: McGraw-Hill. EAGLE, M., WOLITZKY, D. I. & WIN, G. S. (1966). Imagery: effect of a concealed figure in B stimulus. Science, N . Y . 151,837-839. FISHER, C. (1960). Preconscious stimulation in dream, associations and images. Itlonogr. 7: Psychol. I s m s 2, 1-40. PAUL,I. H. & FISHER,c. (1959). S u b l d a l visual stimulation: a study of its influence on subsequent images and dream. J . new. me&. Db. 129, 316-340. H. (1973). Person81 communication. PHILLIPSON, SILVERWAN, L. H. t SILVERMAN, D. K. (1964). A clinical-experimental approach to the study of subliminal stimulation. J . abraorm. SOC.Psychol. 69, 158-172. Soram, D. E. & WILDING, J. M. (1973). Perception without awareness in a dichoptic viewing situation. Br. J . Psychol. 64, 339-349. YARBUS, A. L. (1967). Eye-movements and Viaion. New York: Plenum Preess.
(Manwcript received 29 July 1974; revised mnuscri.pt received 16 September 1975)
