long-term effects of covert face recognition

Brief article

Long-term effects of covert face recognition

Rob Jenkinsa,*, A. Mike Burtona, Andrew W. Ellisb

aUniversity of Glasgow, Glasgow, UKbUniversity of York, York, UK

Received 20 February 2002; accepted 9 August 2002

Abstract

Covert face recognition has previously been thought to produce only very short-lasting effects. In

this study we demonstrate that manipulating subjects’ attentional load affects explicit, but not

implicit memory for faces, and that implicit effects can persist over much longer intervals than is

normally reported. Subjects performed letter-string tasks of high vs. low perceptual load (Lavie, N.

(1995). Perceptual load as a necessary condition for selective attention. Journal of Experimental

Psychology: Human Perception and Perfomance. 21, 451–468.), while ignoring task-irrelevant

celebrity faces. Memory for the faces was then assessed using (a) a surprise recognition test for

the celebrities’ names, and (b) repetition priming in a face familiarity task. The load manipulation

strongly influenced explicit recognition memory, but had no effect on repetition priming from the

same items. Moreover, faces from the high load condition produced the same amount of priming

whether they were explicitly remembered or not. This result resolves a long-standing anomaly in the

face recognition literature, and is discussed in relation to covert processing in prosopagnosia.

q 2002 Elsevier Science B.V. All rights reserved.

Keywords: Face recognition; Covert processing; Priming; Visual attention

1. Introduction

Observers presented with faces that they do not recognize overtly can nevertheless show

signs of having done so covertly. Perhaps the most striking example comes from the study

of prosopagnosic patients, some of whom show evidence of recognizing familiar people

when tested indirectly (for example through measures of galvanic skin response) despite

reporting no overt recognition (e.g. Bauer, 1984; Tranel & Damasio, 1988). Similar results

have been found in semantic priming studies, in which presentation of a face (e.g. Oliver

R. Jenkins et al. / Cognition 86 (2002) B43–B52 B43

Cognition 86 (2002) B43–B52www.elsevier.com/locate/cognit

0010-0277/02/$ - see front matter q 2002 Elsevier Science B.V. All rights reserved.

PII: S0010-0277(02)00172-5

* Corresponding author. Department of Psychology, University of Glasgow, 58 Hillhead Street, Glasgow G12

8QB, UK. Tel.: 144-141-330-3948; fax: 144-141-330-4606.

E-mail address: [email protected] (R. Jenkins).

Hardy’s) typically facilitates recognition of a printed related name (e.g. “Stan Laurel”)

presented immediately afterwards (i.e. within a few seconds of the prime item; Bruce &

Valentine, 1986). Normal subjects show this pattern of semantic priming routinely, but so

do some prosopagnosics, despite not recognizing the prime face (Young, Hellawell, & de

Haan, 1988). Normal subjects can also show covert-only recognition in this situation,

when the prime face is presented subliminally (Morrison, Bruce, & Burton, 2000).

Such short-term effects of covert recognition can be demonstrated relatively easily,

using faces or other classes of stimulus (e.g. printed words). By contrast however, long-

term effects of covert recognition, such as repetition priming, have never been found in the

face domain, despite being commonplace in other domains (e.g. word reading; Beaure-

gard, Benhamou, Laurent, & Chertkow, 1999).

Repetition priming refers to a facilitation in identifying an item due to prior exposure to

that item. In the face domain, typical repetition priming experiments require subjects to

perform some task on faces in a prime phase, perhaps a judgement based on physical

appearance. An unexpected test phase follows, commonly about 20 min later, in which

subjects are asked to make familiarity judgements to faces. RTs to make these judgements

are reliably faster for items that appeared in the prime phase than for items that did not

(e.g. Bruce & Valentine, 1985; Ellis, Young, Flude, & Hay, 1987). Unlike semantic

priming, this effect is domain specific. So, at the intervals typically used, faces prime

faces but not names, and vice versa (Burton, Kelly, & Bruce, 1998; Ellis, Flude, Young, &

Burton, 1996; Ellis et al., 1987). This effect is very robust, and can be seen even when

different photographs of the same face are used at prime and test. Moreover, it survives

radical changes in context between prime and test (e.g. changes in location, stimulus

materials and task; Bruce, Carson, Burton, & Kelly, 1998). Repetition priming from

faces is thus very well explored, and is incorporated into current models of familiar

face recognition (e.g. Burton, Bruce, & Hancock, 1999).

Despite the general robustness of repetition priming for faces, it has never been shown

to occur when stimuli in the prime phase are not overtly recognized. Indeed, Brunas-

Wagstaff, Young, and Ellis (1992) found that even prime faces that were successfully

recognized, but only after prompting (e.g. “he’s a politician…”), did not produce repeti-

tion priming. Thus, it appears that face recognition must be not only overt, but also

spontaneous, in order for repetition priming to occur. Although this effect has survived

in the literature for many years, it does seem anomalous. Why should covert recognition so

easily give rise to immediate semantic priming, but resist longer term repetition priming?

In other domains, such as word reading, unrecognized (subliminal) primes have been

shown to produce both immediate semantic priming (e.g. Hines, Czerwinski, Sawer, &

Dyer, 1986) and longer term repetition priming (e.g. Beauregard et al., 1999).

In the present study we sought to induce long-term covert-only recognition for supra-

liminal faces in normal subjects by manipulating attention to prime faces. According to

Lavie’s perceptual load theory of selective attention (Lavie, 1995, 2000), perceptual proces-

sing is capacity-limited. But within these limits, processing proceeds automatically, from

relevant to irrelevant stimuli, until available capacity is exhausted. The amount of task-

irrelevant processing that takes place is thus influenced by the relevant perceptual load: if

the perceptual load of the relevant task is high enough to exhaust capacity, then irrelevant

stimuli will not be processed. On the other hand, if the relevant perceptual load is low, spare

R. Jenkins et al. / Cognition 86 (2002) B43–B52B44

capacity will inevitably ‘spill over’ to the processing of task-irrelevant stimuli. There is now

considerable evidence in support of Lavie’s theory, from both psychological studies (e.g.

Lavie, 1995, 2000) and neuroimaging studies (e.g. Rees, Frith, & Lavie, 1997). The major

finding, which holds true for various manipulations of perceptual load, is that task-irrelevant

processing is only found under conditions of low perceptual load in the relevant task, and is

eliminated under conditions of high perceptual load.

To date, however, any influence of relevant perceptual load at prime, on covert memory

for task-irrelevant stimuli presented at prime, has not been examined. If overt and covert

recognition measures can be dissociated (e.g. Beauregard et al., 1999; Morrison et al.,

2000), it is possible that repetition priming might reveal some covert recognition for task-

irrelevant faces that were previously presented under high load, even if no overt memory

for those faces is reported.

Here we tested this possibility by presenting displays composed of a letter-string super-

imposed on a famous face (see Fig. 1). In the low load condition, subjects were asked to

respond to the colour of the letter-strings, a task thought to place minimal demands on

attention (e.g. Treisman, 1993). In the high load condition, subjects had to identify a target

letter among similarly-shaped letters in the string, a task thought to demand focused atten-

tion, and shown in previous studies to reduce on-line distractor processing (Lavie, 1995).

Incidental memory for the task-irrelevant famous faces was then assessed using two

measures. First, overt recognition memory was assessed using an old/new discrimination

task for the celebrities’ names (e.g. “was Bill Clinton presented?”). Second, repetition

priming effects were assessed using a speeded familiarity task for the celebrities’ faces.

We predicted that faces presented under low load would be remembered better than those

presented under high load. If so, it should then be possible to establish whether the faces

that are not overtly recognized can nevertheless give rise to repetition priming.

2. Method

2.1. Design and stimuli

Displays in the selective attention task consisted of a central letter-string superimposed

on a famous face at fixation (see Fig. 1). The letter-strings could be either red or blue, and

consisted of a target letter (X or N) and five other angular letters in a random order. All

face images were greyscale photographs with the background removed.

The faces and printed names of 72 celebrities served as stimuli. Twenty-four faces were

presented in the low load condition, and 24 were presented in the high load condition. The

remaining 24 were used as new items at test. Between subjects, the three face sets were

rotated around experimental conditions (high, low and new), so that when pooling over the

whole experiment, each face appeared in each condition equally often. Novel photographs

of all 72 famous faces, and matched photographs of 72 unfamiliar faces, were used in the

face familiarity task.

2.2. Procedure

The experiment consisted of three stages, which were separated by 5 min intervals.


Stage 1: the selective attention displays were presented for 200 ms. Twenty-four subjects

made speeded keypress responses to the colour of the letter-string (red or blue) in the low

load condition, and to the identity of the target letter (X or N) in the high load condition.

Subjects completed two randomized blocks (either low–high, or high–low), each consist-

ing of 24 trials. Each prime face was thus encountered exactly once. Subjects were

instructed to focus on the letter-strings and to ignore the faces throughout. Stage 2:

after the selective attention task, subjects performed a surprise name recognition test on

all 72 celebrities’ names, responding “yes” to celebrities who had been presented in the

selective attention task, and “no” to celebrities who had not. Stage 3: subjects then used


Fig. 1. Example of the type of display used in Stage 1. Subjects responded to a string of letters superimposed on a

task-irrelevant famous face (non-famous face used in this example for copyright reasons). In the low load

condition, subjects responded to the colour of the letter-string (red vs. blue). In the high load condition, they

responded to the identity of a target letter (X vs. N).

keypress responses to make speeded familiarity judgements to 144 faces (all 72 celebrities,

randomly intermixed with 72 unfamiliars).

3. Results

Perceptual load was successfully manipulated. Mean correct RTs and error rates in the

selective attention task were significantly higher in the high load condition (mean RT 767

ms, 29% errors) than in the low load condition (mean RT 310 ms, 5% errors)

(tð23Þ ¼ 13:4, P , 0:05 for RTs; tð23Þ ¼ 11:5, P , 0:05 for errors). The mean percen-

tages of “yes” responses in the name recognition task are presented in Fig. 2.

One-way ANOVA revealed a significant effect of experimental condition

(Fð2; 46Þ ¼ 69:9, P , 0:05). Planned comparisons showed that, as expected, subjects

responded “yes” more frequently to low load and high load celebrities than to new

celebrities (tð23Þ ¼ 11:8, P , 0:05 and tð23Þ ¼ 6:26, P , 0:05, respectively). More

importantly, there was a significant effect of perceptual load on recognition memory

performance, with more “yes” responses to low load celebrities than to high load celeb-

rities (tð23Þ ¼ 5:54, P , 0:05).


Fig. 2. Mean percentage of “yes” responses (n ¼ 24) in the surprise recognition test for famous names (Stage 2).

Overt recognition performance is shown as a function of experimental condition; low load, high load, or new

(foils).

Mean correct RTs and error rates in the face familiarity task are presented in Fig. 3, as a

function of experimental condition.

ANOVA on these RT data showed a significant effect of experimental condition

(Fð2; 46Þ ¼ 6:99, P , 0:05). Planned comparisons confirmed that subjects responded

faster to low load faces and high load faces than to new faces (tð23Þ ¼ 3:23, P , 0:05

and tð23Þ ¼ 3:25, P , 0:05, respectively), indicating repetition priming. However, as Fig.

3 shows, the size of this priming effect was unaffected by the level of perceptual load (43

ms for both load conditions). Error rates were consistent across conditions (15 ^ 1%), and

were not analyzed further.

To provide a more stringent test of dissociation between repetition priming and explicit

recognition memory, priming effects were also analyzed separately for celebrities eliciting

“yes” vs. “no” responses in the overt recognition memory task. Three subjects with empty

cells were excluded from this analysis. Mean correct RTs for the remaining 21 subjects are

shown in Fig. 4.

ANOVA conducted on these split RT data revealed a significant effect of experimental

condition for both the “yes” subset (Fð2; 40Þ ¼ 4:03, P , 0:05), and the “no” subset


Fig. 3. Mean correct RTs (n ¼ 24) in the face familiarity task (Stage 3). Covert recognition performance is shown

as a function of experimental condition; low load, high load, or new (foils).

(Fð2; 40Þ ¼ 3:60, P , 0:05). However, planned comparisons found no significant effect of

load for either subset (t , 1 for the “yes” subset, and tð1; 20Þ ¼ 1:91, n.s. for the “no”

subset). Most importantly, high load faces produced the same amount of priming (50 ms),

regardless of whether or not they were explicitly remembered. Low load faces produced

more priming when they were explicitly remembered (58 ms) than when they were not (13

ms) (tð1; 20Þ ¼ 2:22, P , 0:05).

4. Discussion

This experiment demonstrates, for the first time, a dissociation between overt recogni-

tion memory and repetition priming for faces. Specifically, incidental overt recognition

(Stage 2), for faces that that were irrelevant when first presented, was strongly affected by

the perceptual load of the relevant task at the time of face encoding (Stage 1), with poorer

performance on faces presented under high relevant load. However, this load manipulation

had no effect on repetition priming from the same faces (Stage 3). Note that the priming

observed at Stage 3 cannot have been influenced by the explicit memory test at Stage 2;

although the explicit name recognition test always preceded the face familiarity test, it is


Fig. 4. Mean correct RTs (n ¼ 21) in the face familiarity task (Stage 3) as a function of experimental condition

(low load, high load, or new), shown separately for celebrities eliciting “yes” vs. “no” responses in the name

recognition test (Stage 2).

well-established that repetition priming onto a familiarity decision does not cross stimulus

domains (e.g. names do not prime faces; see Section 1). In any case, since all 72 celebrities

were presented at Stage 2 and at Stage 3, even priming from the names could not explain

the differential priming effect seen here. Instead, it appears that faces presented as irrele-

vant stimuli during a high load relevant task are nevertheless processed by the face

recognition system. Moreover, this processing is evidently relatively deep, involving

access to the person’s identity, since it survives a change in image between prime and

test. Given that perceptual load has previously been shown to modulate processes such as

on-line distractor processing (e.g. Lavie, 1995) and even motion-related activity in early

visual cortex (Rees et al., 1997), the depth of covert processing for the high load faces here

is quite striking. However, it is consistent with reports from other domains (e.g. word

reading) of repetition priming from unattended stimuli (e.g. Kellogg, Newcombe,

Kammer, & Schmitt, 1996; see also Jacoby, Toth, & Yonelinas, 1993, for similar disso-

ciations between overt and covert measures).

How do the present results relate to the Brunas-Wagstaff et al. (1992) finding that

spontaneous recognition is needed for repetition priming? The data in Fig. 4, which

show no significant priming from unrecognized celebrities from the low load condition,

may provide a clue here. It is commonly found that only familiar faces produce repetition

priming, and that this effect arises in the part of the system responsible for retrieving

identity (e.g. Ellis et al., 1987, 1996; but see Goshen-Gottstein & Ganel, 2000, for recent

claims of repetition priming from unfamiliar faces). It also seems likely that in any set of

‘celebrities’, there may be a few who are actually unknown to some subjects. Clearly,

unknown people cannot be overtly recognized by their names, even if their faces have been

fully processed. Thus, it seems plausible that at least some of the ‘unrecognized’ celeb-

rities from the low load condition were rejected at Stage 2 simply because they were

unknown. However, celebrities from the high load condition could be rejected for two

reasons; either because they were unknown, or because there were insufficient processing

resources available at the time of encoding (due to the experimental manipulation).

Indeed, the finding that rejected celebrities from the high load condition nevertheless

produced priming suggests that at least some of those rejected celebrities were known

to the subjects, and hence produced facilitation. We suggest that the results of Brunas-

Wagstaff et al. (1992) may be analogous. It may be that the faces in their study that

required prompting for recognition were actually rather poorly known by the subjects.

If so, one could explain their result not in terms of the spontaneous/prompted dimension

currently thought to be critical, but in terms of the degree of familiarity of the faces. Future

studies could test this directly by deliberately manipulating the familiarity of the prime

faces.

Finally, we return to the issue of covert recognition in prosopagnosia, which has been

the focus of most research in covert face processing (e.g. Burton, Young, Bruce, Johnston,

& Ellis, 1991; Farah, O’Reilly, & Vecera, 1993; Young & Burton, 1999). One important

aspect of covert face recognition is that it may play a role in mediating provoked overt

recognition, a phenomenon whereby some prosopagnosics can achieve transient relief

from their condition, given appropriate cueing (DeHaan, Young, & Newcombe, 1991;

Sergent & Poncet, 1990). Thus, a direct method for studying the interplay between covert-

only and normal overt face recognition may provide a significant step towards treatment.


To date, such a method has remained elusive, partly because covert-only recognition for

supraliminal faces has never previously been demonstrated in neurologically intact

subjects. In the present paper we have provided a method for inducing long-term

covert-only face recognition in normals. This technique provides a platform for further

experiments aiming to tap these effects which could inform the development of rehabilita-

tion strategies for prosopagnosic patients.

Acknowledgements

This work was supported by an ESRC grant to A.M. Burton and A.W. Ellis (R000 23

8246).

References

Bauer, R. M. (1984). Autonomic recognition of names and faces in prosopagnosia: a neuropsychological applica-

tion of the guilty knowledge test. Neuropsychologia, 22, 457–469.

Beauregard, M., Benhamou, J., Laurent, C., & Chertkow, H. (1999). Word priming without awareness: a new

approach to circumvent explicit memory contamination. Brain and Cognition, 39, 149–169.

Bruce, V., Carson, D., Burton, A. M., & Kelly, S. (1998). Prime time advertisements: repetition priming from

faces seen on subject recruitment posters. Memory & Cognition, 26, 502–515.

Bruce, V., & Valentine, T. (1985). Identity priming in the recognition of familiar faces. British Journal of

Psychology, 76, 373–383.

Bruce, V., & Valentine, T. (1986). Semantic priming of familiar faces. Quarterly Journal of Experimental

Psychology, 38A, 125–150.

Brunas-Wagstaff, J., Young, A. W., & Ellis, A. W. (1992). Repetition priming follows spontaneous but not

prompted recognition of familiar faces. Quarterly Journal of Experimental Psychology, 44A, 423–454.

Burton, A. M., Bruce, V., & Hancock, P. J. B. (1999). From pixels to people: a model of familiar face recognition.

Cognitive Science, 23, 1–31.

Burton, A. M., Kelly, S. W., & Bruce, V. (1998). Cross-domain repetition priming in person recognition.

Quarterly Journal of Experimental Psychology, 51A, 515–529.

Burton, A. M., Young, A. W., Bruce, V., Johnston, R. A., & Ellis, A. W. (1991). Understanding covert recogni-

tion. Cognition, 39, 129–166.

DeHaan, E. H. F., Young, A. W., & Newcombe, F. (1991). Covert and overt recognition in prosopagnosia. Brain,

114, 2575–2591.

Ellis, A. W., Flude, B. M., Young, A., & Burton, A. M. (1996). Two loci of repetition priming in the recognition

of familiar faces. Journal of Experimental Psychology: Learning, Memory, and Cognition, 22, 295–308.

Ellis, A. W., Young, A. W., Flude, B. M., & Hay, D. C. (1987). Repetition priming of face recognition. Quarterly

Journal of Experimental Psychology, 39A, 193–210.

Farah, M. J., O’Reilly, R. C., & Vecera, S. P. (1993). Dissociated overt and covert recognition as an emergent

property of a lesioned neural-network. Psychological Review, 100, 571–588.

Goshen-Gottstein, Y., & Ganel, T. (2000). Repetition priming for familiar and unfamiliar faces in a sex-judgment

task: evidence for a common route for the processing of sex and identity. Journal of Experimental Psychol-

ogy: Learning, Memory, and Cognition, 26, 1198–1214.

Hines, D., Czerwinski, M., Sawyer, P. K., & Dwyer, M. (1986). Automatic semantic priming: effect of category

exemplar level and word-association level. Journal of Experimental Psychology: Human Perception and

Performance, 12, 370–379.

Jacoby, L. L., Toth, J. P., & Yonelinas, A. P. (1993). Separating conscious and unconscious influences of

memory: measuring recollection. Journal of Experimental Psychology: General, 122, 139–154.

Kellogg, R. T., Newcombe, C., Kammer, D., & Schmitt, K. (1996). Attention in direct and indirect memory tasks

with short- and long-term probes. American Journal of Psychology, 109, 205–217.


Lavie, N. (1995). Perceptual load as a necessary condition for selective attention. Journal of Experimental

Psychology: Human Perception and Performance, 21, 451–468.

Lavie, N. (2000). Selective attention and cognitive control: dissociating attentional functions through different

types of load. In S. Monsell & J. Driver (Eds.), Control of cognitive processes. Attention and performance (pp.

175–194). Vol. XVIII. Cambridge, MA: MIT Press.

Morrison, D. J., Bruce, V., & Burton, A. M. (2000). Covert face recognition in neurologically intact participants.

Psychological Research, 63, 83–94.

Rees, G., Frith, C. D., & Lavie, N. (1997). Modulating irrelevant motion perception by varying attentional load in

an unrelated task. Science, 278, 1616–1619.

Sergent, J., & Poncet, M. (1990). From covert to overt recognition of faces in a prosopagnosic patient. Brain, 113,

989–1004.

Tranel, D., & Damasio, A. R. (1988). Non-conscious face recognition in patients with face agnosia. Behavioural

Brain Research, 30, 235–249.

Treisman, A. (1993). The perception of features and objects. In A. D. Baddeley & L. Weiskrantz (Eds.), Attention:

awareness, selection, and control. Oxford: Oxford University Press.

Young, A. W., & Burton, A. M. (1999). Simulating face recognition: implications for modelling cognition.

Cognitive Neuropsychology, 16, 1–48.

Young, A. W., Hellawell, D., & de Haan, E. H. F. (1988). Cross-domain semantic priming in normal subjects and

a prosopagnosic patient. Quarterly Journal of Experimental Psychology, 40A, 561–580.


long-term effects of covert face recognition

Documents