Eye movements and visual imagery in free recall

Acta Psychologica 38, 267-276. 0 North-Holland Publishing Company 1974

EYE MOVEMENTS AND VISUAL IMAGERY IN FREE

Wiel H. JANSSEN and Calvin F. NODINE*

Institute for Perception TNO, Soesterherg The Netherlands

RECALL

The relationship between eye movements and visual imagery has almost exclusively been

studied by treating eye movements as the dependent variable while an imagery task is being

performed. in the present experiment three eye-movement treatment conditions were manipu-

lated, within Ss, as the independent variable in order to study their effects on the free recall of

nouns which Ss had to store by means of imagery. The imagery-evoking capacity (I) of nouns

was varied over three levels within lists of Dutch nouns (Low, Medium and High I). Ss were

instructed to generate a visual image to each separate noun under the following treatment

conditions: (a) while they looked over and scanned their image as if they were looking at the

real object; (b) while they received concurrent visual stimulation from a checkerboard pattern; (c) while they fixated on a target. Reliable, but minor effects of treatment conditions on the

recall scores were found. The results were discussed in terms of possible theories about the nature of the relationship between eye movements and visual imagery.

The viewpoint underlying the use of eye-movement measures in visual imagery research is that an image of a stimulus is a reconstruc- tion, though not necessarily a reproduction. of a corresponding percept. Since perception involves scanning the stimulus by movements of the eyes it follows that an image might also be accompanied by similar movements as those displayed on the presence of the stimulus. This reconstructive viewpoint, with its modern-cognitivistic flavor, was ex- pressed long ago by Stricker (1882) and Ladd (1892). and has been rephrased recently by, among others, Rey ( 1958) and Hebb ( 1968).

Two general methods of attack on the relation between eye movements and visual imagery have been followed in the experimental litera- ture.

In most studies, eye movements have functioned as the dependent variable. In these studies Ss have typically been instructed to form images to a set of prespecified stimuli while E observes the eye movc- ments as Ss carry out the imagery task (e.g.. Antrobus ct al. 1964;

* On leave from Temple University, Philadelphia, Penn., USA.

268 W.H. Janssen, CF. Nodine, Eye movements and visual imagery

Brown 1968; Dcckert 1964; Goldthwait 1933; Perky 1910; Schifferli 1953; Totten 193.5). Eye movements have been manipulated as the independent variable in only one study. Hale and Simpson (1970) instructed Ss to generate interactive images linking two given nouns, and to press a button when such an image had emerged, under instructions which required Ss to: (a) make eye movements during the imagery task; (1~) not make eye movements during the imagery task (by fixating on a target), but to think about or imagine making eye movements; (c) not make eye movements nor think about making eye movements during the imagery task. No differences in the ease of generating images, as indicated by response latencies, were found among these treatments. Also, no relationships were found among treatment conditions and vividness ratings of the images as given by the Ss. These findings obtained both when treatments were varied within and between Ss.

The present experiment was designed to assess whether the opportunity to make eye movements influences recall performance when the items to be recalled are stored by means of imagery. The experimental task was free recall of isolated nouns of various imagery-evoking capacity (Low, Medium and High 0. That imagery may indeed be a powerful mediator in free recall has been shown by Paivio et al. (1969). Thus, if the opportunity to make eye movements influences the ease of forming an image, this should reflect itself in the recall scores. Three treatment conditions provided varying opportunities to make eye movements during imaging. In one, Ss were encouraged to make eye movements as if they were looking at a real object. In the second, concurrent visual stimulation during imagery was provided by having the Ss look at a checkerboard pattern. In the remainin, ~7 condition, Ss had to fixate a target while imaging. All Ss were given repeated presentations of noun lists in order to follow the development of performance over trials under different treatments.

Method

Subjects

The Ss were 18 Utrecht University students chosen from the Institute’s pool of regular Ss.

W.H. Janssen, C.F. Nodine, Eye movements and visual imagery 269

Experimental design

There were three treatment conditions: (a) the Ss were instructed to look over and scan

their images freely as if they were looking at the real object (FREE condition); (b) the Ss wcrc

instructed to inspect a checkerboard pattern while imaging (CHECK condition); (c) the Ss were

instructed to fixate on a target while imaging (FIX condition). Treatment conditions were

varied within Ss. Under all treatment conditions Ss were instructed to form a visual image to

each separate noun that was presented.

Stimdus material

Three equivalent experimental lists of 24 nouns each were constructed by consulting a list

of I-values of Dutch nouns (Janssen 1973). The lists were equated for average I, frequency of

occurrence of nouns in the language, and word length, both in terms of lcttcrs and syllables.

Within each list there were 8 Low I. 8 Medium I, and 8 High I nouns. Average I-values wcrc 2.2,

4.5, and 6.6, respectively, with ranges 1.5 -2.5,4.0-5.0, and 6.2-6.9. An equivalent fourth list

was constructed for use in training only. Of the three experimental lists four versions wcrc taperecorded, with a different random word order in each version. \Vords were spoken by a

male speaker at a rate of 1 every 2 set, and wcrc pronounced within a 1 see interval. Six see

prior to the first word of a list a 1000 cps 0.5 set warning tone was presented, and the list wa\ concluded by the same tone at the instant when the 25th word would have occurred.

Apparatus

Horizontal eye movements were monitored by means of electrodes attached to the outer

canthus of each eye. The ground electrode was attached to the center of the forehead. The

EOG’s were amplified and recorded on one of the channels of a Watanabe oscillograph. In order to minimize movement artifacts S was seated in a dental chair with his head resting in a head

clamp. A slide projector located above and behind S’s head provided a uniformly lighted field

of vision on a projection screen in an otherwise dark room. The screen was approximately

1.5 m in front of the Ss, and its dimensions were approximately 80 X 80 deg. In the FIX

condition a removable target of 1 deg diameter was fastened to the center of the screen. In the

CHECK condition a checkerboard pattern, with 3 X 3 deg black and white squares, was pro-

jected onto the screen. In the FREE condition only the blank field was shown. The tape

recorder that presented stimuli to the S was coupled to one channel of the oscillograph, so that

stimuli were graphically represented on the chart paper. A second tape recorder recorded S’s

oral recall.

Procedure

The Ss were tested individually. After applying the electrodes and calibrating the apparatus S received his instructions. He was told that he would hear series of 24 nouns and that he would

have to form an appropriate image to each word, even though this might be difficult in some

instances. (As examples the E gave the words ‘chair’ and ‘communism’.) S was told that he

would have to recall as many words as possible, in any order, within a 90 set interval following the presentation of a list. Order of treatments and lists within treatments was counterbalanced

across the 18 Ss.

270 W.H. Janssen, C.F. Nodine, Eye movements and visual imagery

Each list was presented four times, with a different word order on each trial. A trial with the

training list under the appropriate treatment condition preceded the experimental trials. Differ-

ent treatment conditions were run in separate sessions. Two Ss were tested alternately. While

one S was being tested, the other paused.

Recalibration of the apparatus took place at the start of each treatment session, and also

between trials.

Results

Recall performaxe

Recall scores were obtained from the taperecordings made during the experimental sessions.

A 3 X 3 X 4 analysis of variance (ANOVA) compared the three treatments, the three types of

nouns, and the four trials, on the mean recall scores.’ The ANOVA showed a highly significant

effect of trials (F = 328.0, df= 3/Sl, p < 0.0001) which accounted for 40.7% of the total

non-error variance (see Riemersma and Burry 1973, for the computation of percentages of

variance accounted for by the factors in an ANOVA). There also was a highly significant effect

of I (F = 21 .O, df= 2/34. p = 0.0002), accounting for 5.8% of the total noncrror variance. By a

Newman-Keuls test no difference in recall was found between High and Medium I nouns, but

both were significantly different from Low I nouns.

The main effect of treatments was not significant (E’ = 1 SO. df = 2/34, p = 0.237). llt-rwcver,

two interactions involving treatments had small but reliable effects. These were the treatment

by trial interaction (F= 2.17, df= 6/102, p = 0.051) accounting for 0.3% variance, and the

treatment by I interaction (F = 2.46, df= 4/68, p = 0.053) which accounted for 0.6% variance.

The appropriate means arc shown graphically in figs. 1 and 2.

The significant treatment by trial interaction (fig. 1) was due to differential effects of

treatments on the first trial only, with the FREE condition leading to a higher recall score than

both the other conditions (between which there was no difference in recall). The treatment by I

interaction (fig. 2) indicated that the FREE condition improved recall to lliyh and Medium I

nouns more. relative to Low I nouns, than the other treatment conditions.

The l<OG records wcrc inspected to determine how many cyc movement? had occurred

under different treatments. and in particular. to cheek whether Ss had comphcd to the instruc-

tions in the 1:1X condition. An eye movcmcnt was counted whencvcr there was a dcilection of

at lcaqt 3 deg followed by a fisation of at lcxst 0.2 set duration. I:ye movements were counted over the period from the beginnin, 17 of the first word to the end of the list as indicated by the tune.

On the average, 1 .2 cyc movements per S were made during a list presentation (a trial) in tllc

‘Since there were a few missing data (12 out of 648 data points, not all with the same S or

under the same treatment condition), due to failures of the recordin g apparatu\, the dfs in the

ANOVA are slightly different from what would be expected on the basis of the experimental

design.

W.H. Jamsen, C.F. Nodine, Eye movements and visual imagery 271

I 1

80 -

70 -

LO - _ FREE

d p---d CHECK .-. FIX

30

1 1 2 ,3 L

TRIALS

Fig. 1. Recall performance over trials as a function of treatment conditions

- FREE a----o CHECK .-. FIX

LOW MEDIUM HIGH

I

Fig. 2. Recall performance as a function of treatment conditions and I-value of nouns.

FIX condition; 5 Ss did not make a single movement, and the worst S made an average of 5.5

eye movements per trial. In the FREE condition 5s made an average of 6 1.4 eye movements per

trial, the range being from 27.3 to 109.8. In the CHECK condition the average was 54.1 per

trial, with a range from 8.5 to 108.5. The difference between the FREE and CHECK condition on the one hand and the FIX condition on the other was, of course, clearly significant (t = 9.61

and 8.50, df= 17, p < 0.001). The FREE and CHECK conditions differed at the 0.10 level

(t = 1.74, df = 17), indicating that the attempt to introduce concurrent visual stimulation in

order to reduce eye movements was reasonably successful.

Product-moment correlations were computed over Ss, for the FREE and CHECK treat-

ments, between the number of eye movements made during presentation of a noun list and the

score on subsequent recall. These correlations were rather low (from -0.09 to +0.26, df = 17),

but all except one were positive. Since these correlations were computed across Ss, however, it was decided to examine the eye movement data somewhat more closely in order to investigate

the possibility that there nevertheless might be underlying relationships between number of eye movements made and performance. This was done only for the FREE condition, as it was

212 W.H. Jansseq CF. hlodine, Eye movements and visual imagery

r I

-1. .y* x/ - LOW1 ”

- MEDIUM I 0-0 HIGH1

% 1 I 1 I L 3 4

TRIALS

Fig. 3. Number of eye movements made over trials as a function of f-value 01‘ nouns

judged to bc unfeasihlc to separate eye movements due to imaging activity from those due to

inspection of the checkerboard pattern in the CIIICCK condition.

Numbcrc of eye movements to individual nouns in the I:REL‘ condition wcrc dctcrmincd by

considering activity during the I SW silent interval following the prescntstion of a noun. This

could be done because the nouns were rcprcscnted on the chart paper of the oscillogrnph.

I,‘ig. 3 shows the average number of eye movements made to individual nouns over trials as a

function of I. The ANOVA over these data resulted in significant effects of I (I+‘= 5.99.

df= 2/32, p = 0.006. 1.9%) and of trials (F = 2.04. df= 3/48. p = 0.059, 1.5:s). Post hoc analy-

sis confirmed the overall presence of an endspurt on the fourth trial after a decrease in eye

movements from trials 1 to 3.

The relationship bctwcen the amount of eye-movcmcnt activity and recall performance was

analysed, for the I:R111: condition. by splitting up the recall score for each S on each trial in a

part associated with prccedin, ~1 eye movement activity below the median activity during the

trial, and a part associated with activity above the median. Median activity was defined in terma

of the distribution of the number of eye movements made to individual nouns for ;I particular S on a particular trial.

The ANOVA indicated that there were significant effects of trials (E‘= 64.7, df= 3/48, p < 0.001, 49.0%), the amount of preceding eye movements when dichotomized at the mcdlan (F= 10.5, df= l/16. (I = 0.005, 1.8%) and the interaction of these two variables (F= 3.62.

df= 3/48, p = 0.007, 0.6%). The results are shown graphically in fig. 4. from which it is appar-

cnt that the amount of preceding eye movements is related to subsequent recall at all trials

except the fourth.

Discussion

The main conclusion to be drawn from the analysis of the recall scores is that the effects of eye-movement conditions, although statisti- cally reliable, produced only minor differences in performance. Thus, manipulation of the opportunity to make eye movements during imaging has only little consequence on subsequent free recall of nouns.


Clearly, there exist more potent variables which influence the ease of generating an image than control over eye movements. In the present experiment such variables were the imagery-evoking capacity of the nouns, and the number of list repetitions.

Nevertheless, it would be interesting to speculate as to the reasons behind the differences between the present results and those of Hale and Simpson. There is no doubt that the precise wording of the instructions concerning what Ss should do with their eyes during the imagery task plays a crucial role in this respect. Hale and Simpson told Ss simply to move their eyes. This may have led some Ss to make more or less random eye movements which acted to inhibit rather than enhance the formation of an image. This was not the case in the present study, as our instructions presumed that eye movements were a more or less natural accompaniment to the generation of an image. However, it may be repeated that the results of the present experiment nevertheless indicate that differential effects of treatments. when they occur. are small even with specific instructions which focus Ss’ eye movements on the image.

The opportunity to make appropriate eye movements, as jnduccd by instructions, must also have caused the significant treatments by I intcr- action that was found. The pattern that is displayed in this interaction,

1 I 1 I I 1 80 -

70 - 2

= 2 60-

E

$ 50-

:

LO -

30 - I t

r I I I I

1 2 3 1 TRIALS

Fig. 4. Recall performance over trials as a function of the associated amount of eye-movement

activity, split at the median.

274 W.H. Janssen, C.F. Nodine, L’ve movemerlts and visual imager)

Low 1 nouns showing no improved recall in the FREE condition. appears interpretable as follows: it is so hard to generate an image to a Low 1 noun, that the opportunity to make eye movements cannot be of any help (at least within the 1 set interval that was allowed to form an image in the present experiment).

It is puzzling. though, that no reliable differences in recall were found between High and Medium I nouns, especially since they did not overlap in I-value. This suggests either that I is not the most potent variable at work (at least for Dutch Ss), or that the 7-point scale on which the nouns were originally rated for / by the norm sample (Janssen 1973) was not linearly related to imagery-evoking capacity. Especially the first possibility would be difficult to reconcile with present evidence on the prominent functions of mental images in memory (Paivio 1971).

The eye-movement data shown in figs. 3 and 4 invite speculation as to the exact nature of the relationship between eye-movement parameters and visual imagery appearing from the significant interactions obtained. Since these data are basically of a correlational nature, caution should be taken in basing conclusions on them concerning the direction of this relationship; that is, do images come first and subsequently give rise to a certain amount of eye movements (the ‘outflow’ theory advo- cated by Deckert 1964). or do images emerge only after appropriate eye movements have preceded (the reconstructive or ‘inflow’ viewpoint)?

Deciding on this point appears to be no easy matter, especially since there arise complications which are caused by the effect of what proba- bly are motivational or attentional factors on eye-movement activity. These are reflected in the gradual decrease of the number of eye movements from trial 1 to 3 in the FREE condition, and the endspurt occurring when Ss know that they are doing the last trial.

Apart from these effects, there appear to be consistent differences in the number of eye movements made to nouns of various Z-value. The reasoning, in terms of an outflow theory, to account for these would have to be that nouns that differ in I elicit images that differ in latency or vividness, to which the Ss consequently produce different amounts of eye movements in case they have previously been instructed to scan their image. Similarly, it might be argued that the reason why better recall is found for nouns which were accompanied by a larger amount of activity (fig. 4) is that vivid images are more likely to be recalled as well as give rise to more eye movements than less vivid images.


There are several aspects of the results that an outflow theory is unable to account for. First, it fails to explain the position of the curve of the number of eye movements associated with Medium I nouns (fig. 3). Second, it cannot account for the significant interactions in the experiment, in particular, for the treatments by I interaction. Thus, if eye movements were immaterial to the existence of the image itself it is inexplicable why Medium and High I nouns should be best recalled in the FREE condition.

An intermediate standpoint between pure inflow and outflow theories is that making eye movements may be sufficient, although not necessary, to generate an image. Thus, images might be generated with the help of eye movements, or in an other, as yet unspecified equivalent way. Again, however, it is hard to see why Ss did sometimes profit from the opportunity to make eye movements if they could freely have shifted between these two ways of forming an image.

Finally, a pure inflow, or reconstructive theory, supplied with the assumption that the amount of eye movements made is directly reflected in the vividness or clarity of an image, is also non-consistent with some aspects of the data. It cannot explain the finding that, although the number of eye movements differs among nouns of various I in the FREE condition, these nouns are also differentially recalled in the FIX condition, where there is no difference in eye-movement activity. Moreover, it cannot explain why the pattern of the amount of eye movements as a function of I (fig. 3) is not the same as the pattern of recall as a function of I (fig. 2).

In conclusion, we may state that we have obtained evidence that the opportunity to make appropriate eye movements may produce reliable, though minor, facilitating effects on the formation of an image as meas- ured by recall performance. Thus, we have corroborated the reconstructive viewpoint in principle. More precise statements as to the exact nature of the relationship between eye-movement parameters and visual imagery, however, will have to await further experimentation by means of more elaborate paradigms.

References

Antrobus, J.S., J.S. Antrobus and J.C. Singer, 1964. Eye movements accompanying daydream-

ing, visual imagery and thought suppression. J. Abnorm. Sot. Psychol. 69, 244-252.

Brown, B.B., 1968. Visual recall ability and eye movement. Psychophysiology 4, 300-306.

216 W.H. Jar~ssrrr, C.F. Noditzc, b:ve mowmerrts and ~~isual imageu

Deckcrt. C;.tl., 1964. Pursuit eye movcmcnts in the absence of a moving stimulus. Scicncc 143.

1192-1193. Goldthwait. C.. 1933. Relation of cyc movements to visual imagery. Amer. J. Psychol. 46.

IOh- 110. tlalc, S.M. and lI.M. Simpson. 1970. l:ffects of eye movements on the rate of discovery and the

vividness of visual images. Perception and Psychophysics 9. 242- 246.

Iiebb, D.O.. 1968. Conccrninp imagery. I’sychol. Rev. 75,466 -477. Janssen, W.H., 1973. Imagery values of 327 nouns (in Dutch). Ned. Tijdschr. Psycho]. 28,

465-475.

Ladd. (‘.‘I‘., 1892. Contribution to the psychology of visual dreams. &find I, 299-304.

Paivio, A., 197 I. Imagery and verbal processcs. New York: Halt. Rinehart and Winston.

Paivio. A.. J.C. Yuillc and T.B. Rogers. 1969. Noun imagery and mesninefulnccc in free and serial recall. J. l‘xp. Psychol. 79. 5095514.

Perky. C.W., 1910. An cxperim~ntal study of imagination. Amer. J. Psycho]. 21. 422-452.

Rcy, A., 1958. Lcs images mentales en psychophysiologic. Dhlcctica 12, 130 145. Ricmersma, J.B.J. and S. Burry. 1973. A general program for the analysis of variance ot

t‘actorial and hierarchical dcsipns. Soestcrbcrg: The Institute for Perception TNO, Report

1973c4.

Schiffcrli. P.. 1953. I:tudc par enrepistremcnt photographiquc de la mutricitd oculaire dans l’exploration, dltns la reconnaissance et dans la repr&sentation visucllc. Monatschrift fiir Pspchiatrie und Neurologic 126, 64- 118.

Stricker, S., 1882. Studien iibcr die Bewcgungsvorstellunpcn. Wicn.

Tottcn, L., 1935. Eye movements during visual imagery. Camp. Psychol. hlonoyr. II. no. 3.

Documents

Eye movements and visual imagery in free recall