This article was downloaded by: [Bar-Ilan University] On: 19 November 2012, At: 03:35 Publisher: Psychology Press

Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

The Quarterly Journal of Experimental Psychology Section A: Human Experimental Psychology

Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/pqja20

Repetition priming and face processing: Priming occurs within the system that responds to the identity of a face
Andrew W. Ellis , Andrew W. Young & Brenda M. Flude
a b c c a b

University of York, U.K. University of Durham, U.K., U.K.

University of Lancaster, U.K., U.K. Version of record first published: 29 May 2007.

To cite this article: Andrew W. Ellis, Andrew W. Young & Brenda M. Flude (1990): Repetition priming and face processing: Priming occurs within the system that responds to the identity of a face, The Quarterly Journal of Experimental Psychology Section A: Human Experimental Psychology, 42:3, 495-512 To link to this article: http://dx.doi.org/10.1080/14640749008401234

PLEASE SCROLL DOWN FOR ARTICLE Full terms and conditions of use: http://www.tandfonline.com/page/ terms-and-conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution,

reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material.

Downloaded by [Bar-Ilan University] at 03:35 19 November 2012

THE QUARTERLY JOURNAL OF EXPERIMENTAL PSYCHOLOGY, 1990,IZA (3) 495-512

Repetition Priming and Face Processing: Priming Occurs within the System that Responds to the Identity of a Face
Andrew W. Ellis
University of York, U.K.
Downloaded by [Bar-Ilan University] at 03:35 19 November 2012

Andrew W. Young
University of Durham, U.K.

Brenda M. Flude
University of Lancaster, U.K.
A familiar stimulus that has recently been recognized will be recognized a second time more quickly and more accurately than if it had not been primed by the earlier encounter. This is the phenomenon of repetition priming. Four experiments on repetition priming of face recognition suggest that repetition priming is a consequence of changes within the system that responds to the familiarity of a stimulus. In Experiment I , classifying familiar faces by occupation facilitated subsequent responses to the same faces in a familiarity decision task (Is this face familiar or unfamiliar?) but not in an expression decision task (Is this face smiling or unsmiling?) or a sex decision task (Is this face male or female?). In Experiment 2, familiar faces showed repetition priming in a familiarity decision task, regardless of whether a familiarity judgment or an expression judgment had been required when the faces were first encountered. Expression decisions to familiar faces again failed to show repetition priming. In Experiment 3, familiar faces showed repetition priming in a familiarity decision task, regardless of whether a familiarity judgment or a sex judgment had been asked for when the faces were first encountered. Sex decisions to familiar faces again failed to show repetition priming. In Experiment 4, familiarity decisions continued to show repetition priming when a brief presentation time with encouragement to respond while the face was displayed reduced response latencies to speeds comparable to those for sex and expression judgments in Experiments 1 to 3. The results are problematic for theories that propose that repetition priming is mediated by episodic records of previous acts of stimulus encoding.
Requests for reprints should be sent to Andrew Ellis, Department of Psychology, University of York, Heslington, York YO1 5DD, England. This work was supported by ESRC grants C 0023 2323 and XC15250004. We are grateful to the Press Association for help in finding suitable photographs to use as stimuli.

61990 The Experimental Psychology Society

496

ELLIS, YOUNG, FLUDE

When a familiar stimulus such as a word or a face has recently been encountered, it may be responded to more quickly and more accurately on a second presentation than if it had not been seen earlier. This is the phenomenon of repetition priming. Repetition priming of visual word recognition has been observed in a range of experimental paradigms including recognition thresholds (Morton, 1979a; Neisser, 1954), lexical decision (Scarborough, Cortese, & Scarborough, 1977), naming (Feustel, Shiffrin, & Salasoo, 1983) and word fragment completion (Tulving, Schacter, & Stark, 1982). Repetition priming of familiar face recognition has been observed by Bruce and Valentine (1985), Young, McWeeny, Hay, and Ellis (1986a) and Ellis, Young, Flude, and Hay (1987) in tasks that involve familiarity decisions (deciding as rapidly as possible whether a face is familiar or unfamiliar), semantic decisions (deciding the occupation of a familiar person whose face is seen), and naming briefly presented faces. One account of repetition priming holds that repetition priming occurs within cognitive systems that respond to the familiarity or otherwise of a perceived stimulus. There are several variants of this account in existence. In Mortons logogen theory of word recognition, the lasting effect involves a reduction of the threshold of the logogen (recognition unit) for a word (Morton, 1979a, 1979b). Objections have been raised to logogen theory as an account of repetition priming on the grounds that changes in the thresholds of logogens are only short-lived (Jacoby, 1983a; Roediger & Blaxton, 1987). This objection is based, however, on a misunderstanding of logogen theory. Semantic context can cause a temporary lowering of the thresholds of logogens for words related to the context, but the actual firing of a logogen that occurs when a word is perceived causes a permanent lowering of the threshold (Morton, 1969). In the interactive activation model of McClelland and Rumelhart (1981), the lasting effect involves an increase in the resting level of activation of the node for that word. Both these theories hold that the repetition priming effect is ahistoric in the sense that the recognition units carry no episodic record of where or when a primed word was last encountered or what subsequent processing it underwent. Theories of repetition priming of face recognition have similarly proposed that priming occurs within a cognitive system that responds to the familiarity of seen faces. Bruce and Valentine (1985) suggested that priming may occur within face recognition units that contain stored descriptions of the appearance of familiar faces (see Bruce & Young, 1986; Hay & Young, 1982), and Ellis, Young, Flude, and Hay (1987) argued that the representations that mediate recognition and priming of familiar faces might have the properties of distributed memories (cf. McClelland & Rumelhart, 1985). If any of these accounts of repetition priming is correct, then priming should only be observed in tasks where performance requires that a stimulus be identified as familiar. One advantage of faces as stimuli is that subjects can

Downloaded by [Bar-Ilan University] at 03:35 19 November 2012

REPETITION PRIMING AND FACE PROCESSING

497

Downloaded by [Bar-Ilan University] at 03:35 19 November 2012

be asked to make a range of different yet natural decisions regarding the same set of faces, not all of which require that the familiarity or otherwise of the face be known. Subjects can, for example, be asked to judge whether a face is male or female, or smiling or non-smiling. Recent theories of face processing (e.g. Bruce & Young, 1986; Ellis, Young, & Hay, 1987; Hay & Young, 1982) have proposed that such judgments are made by cognitive systems that operate upon perceived faces independently of, and in parallel to, the systems that determine familiarity. This view is supported by both experimental and neuropsychological evidence. Experimental evidence includes the observation that decisions about the emotional expression of a face are unaffected by whether the face is familiar or unfamiliar (Bruce, 1986; Young, McWeeny, Hay & Ellis, 1986b),as can be decisions about the sex of a face (Bruce, Ellis, Gibling, & Young, 1987). Neuropsychological evidence includes the fact that brain-injured prosopagnosic patients may still be able to judge accurately the age, sex, and expression of a face while no longer being able to recognize once-familiar faces (e.g. Bruyer, Laterre, Seron, Feyereisen, Strypstein, Pierrard, & Rectem, 1983; Tranel, Damasio, & Damasio, 1988). The cognitive systems that mediate expression and sex judgments require descriptions of facial expressions and of the structural properties of male and female faces, but we propose that they do not contain stored descriptions of the faces of particular familiar people. Because they do not contain personspecific representations, they should not be prone to priming based on the repetition of individual faces. This prediction is tested in Experiment 1.

EXPERIMENT 1
Experiment I has a two-phase structure that in general terms is common to all three experiments reported here and to many others in the literature on repetition priming. In Phase 1 of the experiment, subjects make decisions about seen faces. After a brief interval, Phase 2 (the test phase) is introduced to the subject as ostensibly a new task, but a task that once again involves making decisions to faces. In Phase 2 the decisions are always speeded decisions whose latency is recorded. Repetition priming is measured by comparing the reaction times in Phase 2 to faces that were or were not previously seen in Phase 1. In Experiment 1 subjects decided in Phase 1 whether faces belonged to politicians, actors, or sportsmen. This is a semantic decision, which requires that the faces be identified and hence requires that the face recognition units for those faces be activated. Some of the faces encountered in Phase 1 recurred later in Phase 2, allowing the possibility of repetition priming of these faces. In Phase 2 different groups of subjects made one of three speeded decisions. These involved deciding whether each face was

498

ELLIS, YOUNG, FLUDE

familiar or unfamiliar (a familiarity decision), deciding whether each face was smiling or unsmiling (an expression decision), or deciding whether each face was male or female (a sex decision). Of these three decisions, only the familiarity decision involves the face recognition units, and so on the recognition unit theory of repetition priming only familiarity decisions should show the effect.

Method
Downloaded by [Bar-Ilan University] at 03:35 19 November 2012

Subjects. Thirty-six undergraduate and postgraduate students took part in the experiment. Each was paid 2. Design and Stimuli. The test faces for all conditions were 24 male, smiling faces comprising 8 politicians, 8 actors, and 8 sportsmen. These 24 faces were rated by 6 independent raters for familiarity on a 7-point scale (from 1 =unfamiliar to 7=very highly familiar). The test faces were then divided into two sets of 12 (Set A and Set B), matched on familiarity (mean familiarity ratings: Set A 6.04; Set B 6.03). Each set contained 4 actors, 4 politicians and 4 sportsmen. For each subject, one set of 12 faces would be primed by being presented in Phase I , and the other set would be unprimed. In Phase 1 (pretraining), the 12 test faces of Set A or Set B were accompanied by 12 filler faces. These were all male smiling faces: 3 politicians, 4 actors, and 5 sportsmen. In Phase 2 (test), all 24 test faces (Set A+Set B) were accompanied by a total of 44 filler faces, which varied according to the nature of the Phase 2 task. Ten new smiling male faces of individuals who were not actors, politicians, or sportsmen occurred as fillers for all three Phase 2 tasks. In addition, the familiarity decision task (familiar/ unfamiliar) included 34 smiling unfamiliar male faces, the sex decision tusk (male/female) included 34 smiling familiar females, and the expression decision tusk (smiling/non-smiling) included 34 unsmiling familiar male faces. The proportion of faces requiring each of the binary responses (familiar/unfamiliar; male/female; smiling/unsmiling) was thus 0.5 for each Phase 2 task. All faces were presented on black-and-white slides with a circular template surrounding the face to obscure clothing and background. The slides were back-projected onto a white screen and subtended a visual angle of approximately 10". Procedure. The nature of Phase 1 (the pretraining phase) was the same for all subjects. Twelve test faces (Set A or Set B) randomly interleaved with 12 filler faces were presented for 5 sec each. The subject's task was to state whether each face belonged to a politician, actor, or sportsman. Half the subjects saw the Set A test faces, and half saw the Set B test faces.

REPETITION PRIMING AND FACE PROCESSING

499

Downloaded by [Bar-Ilan University] at 03:35 19 November 2012

In each Phase 2 task all 24 test faces were presented, randomly interleaved with differing fillers, as described above. In the fumiliuriry decision task the subject was required to decide as quickly as possible whether each face was familiar or unfamiliar; in the sex decision task to decide as quickly as possible whether each face was male or female; and in the expression decision task to decide as quickly as possible whether each face was smiling or unsmiling. The decision was made by pressing one of two buttons situated beneath the index fingers of the left and right hands. Twelve subjects took part in each Phase 2 task. Of these, 6 would have been pretrained on Set A test faces in Phase 1 and 6 on Set B test faces. The Phase 2 tasks were introduced immediately after the Phase 1 task was completed. Eight practice faces preceded the experimental stimuli for each Phase 2 task. The exposure presentation for each face was 2.5 sec, with an interstimulus interval of approximately 5 sec. The interval between a primed test face being seen in Phase 1 and again in Phase 2 varied from a minimum of around 3 min to a maximum of around 15 min. Of the 24 faces seen in Phase I, 12 were repeated in Phase 2, and 12 of the 68 faces seen in Phase 2 (1 2/34 familiar faces) were repeats from Phase 1. Results The error rate in Phase 2 of Experiment 1 was low (4.1 YO), errors were and not analysed. A two-way, by-subjects analysis of variance was carried out on the mean correct reaction times (RTs), with Phase 2 Task (expression, sex, or familiarity decision) as a between-groups factor, and priming as a withinsubjects factor. The results are shown in Table 1. The effect of Phase 2 task was significant, F1(2,33)= 9.30, ~ ~ 0 . 0Post1. hoc Tukey tests (a=0.05) showed that overall RTs in the familiarity condition were slower than reaction times in the expression condition and the sex condition, which did not differ significantly.
TABLE 1 Mean RTs from Phase 2 of Experiment 1 to Familiar Faces Seen or Not Seen in Phase 1 of the Experiment Phase I Task Occupation Decision Phase 2 Task

Faniiliurity Decision Expression Decision

Sex Decision

Primed (seen) Unprimed (nor seen) Mean RTs in msec.

709 862

552 566

636 638

500

ELLIS, YOUNG,FLUDE

Downloaded by [Bar-Ilan University] at 03:35 19 November 2012

The main effect of Priming was also significant, F1 ( I , 33) = 30.17, p < O . O O I , with reaction times to primed targets being faster in Phase 2 than reaction times to unprimed targets. This effect was qualified, however, by a Priming x Phase 2 task interaction, F1(2,33) = 21.82, p < 0.001. Post-hoc Tukey tests (a=O.O5) carried out on this interaction showed that there was a significant difference between primed and unprimed faces when the Phase 2 task involved a familiarity decision, but not when the Phase 2 task involved an expression decision or a sex decision. The generalizability of these findings was confirmed by a by-items analysis of variance, which showed significant effects of Phase 2 task, E!(2,46)=39.40, p<O.OOI and Priming, n(l, 23)= 19.64, p<O.OOl, and a significant Priming x Phase 2 Task Interaction, F2(2,46) = 15.1 1, p <0.001.

Discussion
Experiment 1 found significant repetition priming of familiarity decisions made in Phase 2 to faces that had been semantically classified in Phase 1. Expression and sex decisions to familiar faces in Phase 2 did not show repetition priming. Thus, deciding in Phase 1 that Neil Kinnock is a politician rather than an actor or a sportsman reduces the time needed in Phase 2 to decide that the same photograph of Neil Kinnock is of a familiar face but does not affect the time necessary to decide that it is a smiling face or a male face. This pattern of results is as predicted by the theory of repetition priming outlined above. It could be argued, though, that repetition priming depends upon a degree of congruence between the demands of the Phase 1 and Phase 2 tasks. Semantic classification and familiarity decision both require that a face be recognized as familiar and may have sufficient in common to yield priming, whereas semantic classification may have too little in common with expression and sex processing to yield repetition priming. This possibility was explored in Experiments 2 and 3 by having subjects make expression decisions (Experiment 2) or sex decisions (Experiment 3) twice to the same photographs of familiar or unfamiliar faces.

EXPERIMENT 2
In Experiment 2, subjects made speeded expression decisions (smiling or non-smiling) to a mixture of familiar and unfamiliar faces in Phase 1. In Phase 2 they made either speeded expression decisions or speeded familiarity decisions to faces, some of which had been encountered in Phase 1. The theory that repetition priming occurs within the system that guages the familiarity of a face predicts no priming when the Phase 2 task involves expression decision, regardless of how congruent the Phase 1 and Phase 2 tasks may be.

REPETITION PRIMING AND FACE PROCESSING

501

It is an interesting question as to whether this theory predicts repetition priming of familiarity decisions in Phase 2 of Experiment 2. The answer depends on whether, when making expression decisions about familiar faces in Phase I, subjects automatically recognize the faces as belonging to familiar people. If they do, then repetition priming will be found for the familiarity decisions in Phase 2, even though Phase 1 did not explicitly require subjects to identify seen faces as familiar or unfamiliar.

Method
Downloaded by [Bar-Ilan University] at 03:35 19 November 2012

Subjects. Twenty-four undergraduate and postgraduate students took part in the experiment. Each was paid 2. Design and Stimuli. The test faces for all conditions were 64 male faces comprising 16 smiling familiar faces, 16 unsmiling familiar faces, 16 smiling unfamiliar faces, and 16 unsmiling unfamiliar faces. These were divided into 2 sets of 32 (Set A and Set B), each set containing 8 smiling familiar faces, 8 unsmiling familiar faces, 8 smiling unfamiliar faces and 8 unsmiling unfamiliar faces. All the test faces occurred in Phase 2, which involved either a familiarity decision task or an expression decision task. There were no filler faces in Phase 2. The composition of the test faces in Phase 2 ensured that for both the familiarity decision task and the expression decision task the proportion of faces requiring each of the binary responses (familiar/unfamiliar or smiling/unsmiling) was 0.5. In Phase 1 the 32 test faces of either Set A or Set B were randomly interleaved with 32 fillers (8 smiling familiar faces, 8 unsmiling familiar faces, 8 smiling unfamiliar faces, and 8 unsmiling unfamiliar faces). All faces were shown as black-and-white slides with a circular template surrounding the face to obscure clothing and background. The slides were back-projected onto a white screen and subtended a visual angle of approximately lo". Procedure. In Phase 1 all subjects made speeded expression decisions to the 32 test faces of either Set A or Set B, plus the 32 filler faces described above. The test and filler faces were randomly interleaved. Each face was presented for 2.5 sec, and the subject's task was to indicate by pressing one of two buttons as quickly as possible whether the face was smiling or unsmiling. Half the subjects saw the Set A test faces and half saw the Set B test faces. Sixteen practice faces (4 smiling familiar faces, 4 unsmiling familiar faces, 4 smiling unfamiliar faces, and 4 unsmiling unfamiliar faces) were presented at the start of Phase 1 . In Phase 2, all 64 test faces were presented, with Set A and B faces randomly interleaved. In thefamiliarity decision task the subject was required to decide as quickly as possible whether each face was familiar or unfamiliar;

502

ELLIS, YOUNG, FLUDE

Downloaded by [Bar-Ilan University] at 03:35 19 November 2012

and in the expression decision task to decide as quickly as possible whether each face was smiling or unsmiling. The decision was made by pressing one of two buttons situated beneath the index fingers of the left and right hands. Twelve subjects took part in each Phase 2 task. Of these, 6 would have been pretrained on Set A test faces in Phase 1 and 6 on Set B test faces. The Phase 2 tasks were introduced immediately after the Phase 1 task was completed. Sixteen practice faces (4 smiling familiar faces, 4 unsmiling familiar faces, 4 smiling unfamiliar faces, and 4 unsmiling unfamiliar faces) preceded the experimental stimuli for each Phase 2 task. The exposure presentation for each face was 2.5 sec, with an interstimulus interval of approximately 5 sec. The interval between a primed test face being seen in Phase 1 and again in Phase 2 varied from a minimum of around 3 min to a maximum of around 8 min. Of the 64 faces seen in Phase 1,32 were repeated in Phase 2, and 32 of the 64 faces seen in Phase 2 were repeats from Phase 1. Results The error rate in Phase 2 of Experiment 2 was low (3.5%), and errors were not analysed. A three-way, by-subjects analysis of variance was carried out on the mean correct RTs, with Phase 2 task as a between-groups factor Vamiliarity or expression decision), and face familiarity (familiar or unfamiliar) and priming (primed or unprimed) as within-subjects factors. Mean correct RTs are shown in Table 2.

TABLE 2 Mean Reaction Times from Phase 2 of Experiment 2 to Faces Seen or Not Seen in Phase 1 of the Experiment

Phase 1 Tusk Expression Decision Phase 2 Tusk Familiarity Decision Fumiliur faces
,

Expression Decision

Primed (seen) Unprimed (not seen)

Un fum iliur f i r e s

69 1 827

640
664

Primed (seen) Unprimed (not seen) Mean reaction times in msec.

92 I 840

667 653

REPETITION PRIMING AND FACE PROCESSING

503

Downloaded by [Bar-Ilan University] at 03:35 19 November 2012

The main effect of Phase 2 Task was significant, F1(2,22) = 6.47, p < 0.05, with expression decisions being faster than familiarity decisions. The main effect of face Familiarity was also significant, F1(1,22) = 10.77,p CO.01, with overall faster RTs to familiar faces than to unfamiliar faces. This effect was qualified by a face Familiarity x Phase 2 Task interaction, F1(1,22) = 8.28, p < 0.01. Post-hoc Tukey tests ( a = 0.05) showed that this interaction arose because RTs were faster to familiar than to unfamiliar faces in the familiarity decision condition of Phase I1 but not in the expression decision condition. The overall effect of priming and the Priming x Phase 2 Task interaction were not significant. There was, however, a significant face Familiarity x Priming interaction, F1( 1,22) = 20.83, p < 0.01. Post-hoc Tukey tests (a = 0.05) showed that this interaction arose because overall RTs to familiar faces were faster in the primed than in the unprimed condition, whereas RTs to unfamiliar faces were slower in the primed than in the unprimed condition. This interaction was further qualified by a significant 3-way face Familiarity x Priming x Phase 2 Task interaction, F1( 1,22) = 10.45, p c 0.01. Post-hoc Tukey tests (a= 0.05) showed that the pattern just described of faster RTs to primed than unprimedfamiliar faces but slower RTs to primed than unprimed unfamiliar faces holds for familiarity judgments in Phase 2, but not for expression judgments. We consider this 3-way interaction to be the most important result of Experiment 2. All of these findings were confirmed in a by-items analysis, which found significant effects of Phase 2 task, F2( 1,62) = 164.73, p < 0.001, and face Familiarity, F2( 1,62) = 22.94, p < 0.001, in the absence of a significant effect of priming. The two-way face Familiarity x Phase 2 Task interaction was significant, F2( 1,62) = 16.51, p < 0.001, as was the Familiarity x Priming interaction, F2( 1,62) = 14.27, p < 0.001. The three-way Familiarity x Priming x Phase 2 Task interaction was also significant, F2(1,62) = 17.50, p < 0.001.

Discussion
Experiment 2 found no repetition priming of expression decisions, even when both Phase 1 and Phase 2 tasks involved expression decisions to the same photographs of familiar people. Expression decisions in Phase 2 were also unaffected by face familiarity, being as fast to unfamiliar as to familiar faces. This replicates the observation of Bruce (1986) and Young et al. (1986b) that emotional expression decisions to faces are unaffected by the familiarity of the face. Repetition priming was observed for familiar faces when the Phase 2 task required familiarity decisions, even though the Phase 1 expression decision task did not demand that the familiar faces be recognized. This suggests that familiar faces are recognized automatically, irrespective of the task demands,

504

ELLIS, YOUNG, FLUDE

Downloaded by [Bar-Ilan University] at 03:35 19 November 2012

and that such automatic recognition is sufficient to yield repetition priming. This point is further discussed in the General Discussion. Familiarity decisions were slower to repeated than to new unfamiliar faces in the Phase 2 familiarity decision task. This finding was also made by Young et al. (1986a, Experiment 4). It parallels the observation sometimes made that rejection of repeated non-words in a lexical decision task (Is this letter string a word or not?) can be slower than rejection of new non-words (see Feustel, ShiRrin, & Salasoo, 1983). A likely explanation is provided by Humphreys, Besner, and Quinlan (1988), who note that although repetition may under some circumstances facilitate the perception of non-words, the slight feeling of familiarity engendered by encountering a non-word for a second time may delay its rejection as a non-word. Similarly, the slight feeling of familiarity engendered by encountering an initially unfamiliar face for the second time in the space of a few minutes may delay its rejection as unfamiliar. We would note that this effect need not be episodic in nature: it may be that one encounter with an unfamiliar word is sufficient to create a representation in the face recognition system of the sort that, given more encounters, would develop into a lasting face recognition unit.

EXPERIMENT 3
Experiment 3 is similar to Experiment 2, and its rationale is much the same. Experiment 3, however, compares the effect of repetition on familiarity decisions and sex decisions. As in Experiment 2, Phases 1 and 2 were made as comparable as possible by having both involve speeded familiarity or sex decisions to faces. All four combinations of familiarity and sex decision tasks in Phase 1 and Phase 2 were employed in a between-groups comparison. On the basis of the results of Experiment 2, we take the prediction to be that repetition priming of familiar face recognition will be observed whenever the Phase 2 task involves familiarity decisions. Whether the amount of repetition priming will be the same when Phase 1 involves sex decisions as when it involves familiarity decisions is an open and interesting question. Method
Subjects. Forty-eight undergraduate and postgraduate students took part in the experiment. Each was paid f 2 . Design and Stimuli. The test stimuli for all conditions were 64 faces comprising 16 familiar male faces, 16 familiar female faces, 16 unfamiliar male faces, and 16 unfamiliar female faces. For Phase 1 these were divided into 2 sets (Sets A and B) of 32 faces, with each set containing 8 faces of each type. In Phase 1 the 32 faces of Set A or Set B were randomly interleaved with

REPETITION PRIMING AND FACE PROCESSING

505

Downloaded by [Bar-Ilan University] at 03:35 19 November 2012

a further 32 filler faces (8 familiar males, 8 familiar females, 8 unfamiliar males, and 8 unfamiliar females). The stimuli presented to all subjects in all conditions in Phase 2 were the 64 test faces, randomly interleaved and without fillers. The composition of the test faces in Phase 2 ensured that for both the familiarity decision task and the sex decision task the proportion of faces requiring each of the binary responses (familiar/unfamiliar or male/ female) was 0.5. All faces were shown as black-and-white slides with a circular template surrounding the face to obscure clothing and background. The slides were back-projected onto a white screen and subtended a visual angle of approximately lo".
Procedure. Both Phase 1 and Phase 2 involved making speeded responses to faces, though only Phase 2 RTs were analysed, and subjects were required in both Phase 1 and Phase 2 to make either familiarity or sex decisions to the faces. There were four groups of subjects, with 12 subjects in each group. The first group made familiarity decisions in both phases, the second group made sex decisions in both phases, the third group made familiarity decisions in Phase 1 and sex decisions in Phase 2, and the fourth group made sex decisions in Phase 1 and familiarity decisions in Phase 2. The exposure presentation for each face was 2.5 sec, with an interstimulus interval of approximately 5 sec. The decision was made by pressing one of two buttons situated beneath the index fingers of the left and right hands. Each phase was preceded by 16 (different) practice faces (4 familiar males, 4 familiar females, 4 unfamiliar males, and 4 unfamiliar females). The interval between a primed test face being seen in Phase 1 and again in Phase 2 varied from a minimum of around 3 min to a maximum of around 8 min. Half of the 64 faces seen in Phase 1 were repeated in Phase 2, and half of the 64 faces seen in Phase 2 were repeats from Phase 1.

ResuIts
The error rate in Phase 2 of Experiment 3 was low (2.6%), and errors were not analysed. A four-way, by-subjects analysis of variance was carried out on the mean correct RTs from Phase 2, with Phase 1 task vamifiarity or sex decision) and Phase 2 task (also familiarity or sex decision) as betweengroups factors, and Priming (primed or unprimed) and face Familiarity Vamiliar or unfamiliar) as within-subjects factors. All four main effects were significant. The effect of Phase 1 task, F1(1,44) = 4.17, p < 0.05, showed faster overall RTs in Phase 2, when the Phase 1 task had involved a familiarity decision than when it had involved a sex decision. There were no interactions involving Phase 1 task: that is, the

506

ELLIS, YOUNG, FLUDE

TABLE 3 Mean RTs from Phase 2 of Experiment 3 to Faces Seen or Not Seen in Phase 1 of the Experiment

Phase 1 Task Familiariry Decision Phase 2 Task Familiarity Decision Sex Decision Familiarily Decision
Sex

Sex Decision

Decision

Downloaded by [Bar-Ilan University] at 03:35 19 November 2012

Familiar faces Primed (seen) Unprimed (not seen) Unfamiliarfaces Primed (seen) Unprimed (not seen) Mean RTs in msec. 863 84 1 514
581

640 Ill

564 515

616 I11

525 530

190 164

510 511

pattern seen in Phase 2 occurred irrespective of whether the Phase 1 task involved familiarity or sex decisions. In Table 3 the results of Experiment 3 are collapsed across this factor. Themain effect of Phase 2 task, F1(1,44)=52.10,p<0.001, showed faster overall RTs when Phase 2 involved sex decisions than when it involved familiarity decisions. The main effect of face Familiarity, F1(1,44)= 37.53, p < 0.001, showed faster overall RTs to Familiar than unfamiliar faces in Phase 2, and the main effect of Priming, F1(1,44) = 46.89, p < 0.001, showed faster overall RTs to primed than unprimed faces in Phase 2. The factors of Phase 2 task, face Familiarity, and Priming all interacted with one another. Phase 2 Task interacted with face Familiarity, F1( 1,44) = 39.63, p < 0.001. Post-hoc Tukey tests (a= 0.05) showed that RTs to familiar faces were faster than RTs to unfamiliar faces when the Phase 2 task involved a familiarity decision, but not when it involved a sex decision. Phase 2 Task interacted with Priming, P1(1,44)=21.31, p<O.OOl. Post-hoc Tukey tests (u=O.O5) showed faster RTs to primed than unprimed faces when the Phase 2 task involved a familiarity decision, but not when it involved a sex decision. Finally, face Familiarity interacted with Priming, F1(1,44) = 30.33, p < 0.001. Post-hoc Tukey tests (01 = 0.05) showed faster RTs to primed than unprimed familiar faces, but no difference between RTs to primed and unprimed unfamiliar faces. The three-way interaction between Phase 2 Task, face Familiarity, and Priming that is implied by the above pattern of two-way interactions was also

REPETITION PRIMING AND FACE PROCESSING

507

Downloaded by [Bar-Ilan University] at 03:35 19 November 2012

significant, Fl(l,44)=35.57, p < O . O O I . That is, when the Phase 2 task involved a familiarity decision, RTs were faster to primed than unprimed familiar faces, although there was no effect of priming on unfamiliar faces. When the Phase 2 task involved a sex decision, there was no effect of either familiarity or priming. The by-items analysis confirmed the significant main effects of Phase 1 task, n( 1,62)= 2 1 1 36, p < 0.001, Phase 2 task, n( 1,62)= 55 1.70, p < 0.00 1, face Familiarity, F2( I , 62) = 24.94, p < 0.01, and Priming, F2( 1, 62) = 14.85, p < 0.01. All the interactions that were significant in the by-subjects analysis were also significant in the by-items analysis, specifically Familiarity x Priming, n ( 1 , 6 2 ) = 20.05, p<O.OOI, Familiarity x Phase 2 Task, n( = 59.26, p < 0.001, Priming x Phase 2 Task, n( = 7.3 1, 1,62) 1,62) p < 0.01, and Phase 2 Task x Familiarity x Priming, F2( I , 62) = 29.51, p < 0.001.

Discussion
Experiment 3 found repetition priming of familiarity decisions to familiar faces that was as great when the first encounter required a sex decision as when it involved a familiarity decision. In contrast to Experiment 2, there was no significant slowing of responses to repeated unfamiliar faces in the Phase 2 familiarity decision task. This attests to the fragility of the effect whereby a single encounter with an unfamiliar stimulus can leave a trace sufficient to engender a feeling of familiarity when the stimulus is encountered for a second time some minutes later, and retard the rejection of that stimulus as unfamiliar. Sex decisions showed no repetition priming in Phase 2, even when the Phase 1 task involved speeded sex decisions to the same photographs as recurred in Phase 2. There was also no effect of face familiarity on the time taken to decide whether a face was male or female. Bruce (1986) found an effect of familiarity on sex decisions but argued that this was because her set of faces contained some that were particularly hard to classify on physical features alone, so subjects made the sex decisions to some of the familiar faces by first recognizing who the faces belonged to. This interpretation was supported by the results of Bruce, Ellis, Gibling, and Young (1987). The faces employed in Experiment 3 were chosen to be unambiguously male or female and showed no influence of familiarity on sex decision latencies.

EXPERIMENT 4
Experiments 1 through 3 have been consistent in finding repetition priming for familiarity decisions but not for sex or expression decisions. It must be noted, though, that in those experiments familiarity decisions are consis-

508

ELLIS, YOUNG, FLUDE

tently slower than sex or expression decisions. This confounding of type of decision with overall decision speed raises the possibility that repetition priming is only shown by, or is only detectable for, slower decisions to faces. Experiment 4 examines this possibility by using a 600-msec stimulus presentation time combined with an instruction to subjects to respond within the presentation time.

Method
Downloaded by [Bar-Ilan University] at 03:35 19 November 2012

Subjects. Twelve undergraduate and postgraduate students took part in the experiment. Each was paid f2. Design and Stimuli. The Phase 2 test stimuli were 64 faces comprising 32 highly familiar faces and 32 unfamiliar faces, randomly interleaved and without fillers. For Phase 1 the 32 familiar faces were divided into 2 sets (Sets A and B) of 16 faces. In Phase 1 the 16 faces of Set A or Set B were randomly interleaved with a further 16 familiar and 32 different unfamiliar filler faces. Thus, half the famous faces seen in Phase 2 had been seen previously in Phase 1. All faces were shown as black-and-white slides with a circular template surrounding the face to obscure clothing and background. The slides w r e back-projected onto a white screen and subtended a visual angle of approximately 5". Procedure. Both Phase 1 and Phase 2 involved making speeded familiarity decisions to faces, although only Phase 2 RTs were analysed. The exposure duration for each face was 600 msec with an interstimulus interval of approximately 4sec. The decision was made by pressing one of two buttons situated beneath the index fingers of the left and right hands. In both phases subjects were instructed to respond before the slide disappeared from the screen (i.e. within 600msec); RTs longer than 600msec were not, however, treated as errors. Each phase was preceded by 16 (different) practice faces.

Results
A one-way by-subjects analysis of variance on the mean correct RTs to familiar faces in Phase 2 showed a significant effect of conditions, F( 1, 1 1) = 8 1.65, p < 0.001, with RTs to primed faces (5 19 msec, error rate 1.6%) being faster than to unprimed faces (598 msec, error rate 10%). The effect of conditions was also significant on a by-items analysis, F(1,31)=33.13, p < O . O O l .

REPETITION PRIMING AND FACE PROCESSING

509

Discussion
The introduction into Experiment 4 of brief presentations and a response deadline successfully reduced familiarity decision times to the same range as sex and expression decisions in Experiments 1 to 3, yet familiarity decisions continued to show repetition priming where sex and expression decisions had not. This cross-experiment comparison was supported by a two-way analysis of variance, which combined the familiarity decision latencies from Experiment 4 (primed 519 msec; unprimed 598 msec) with the expression decision latencies from Experiment 2 (primed 640 msec; unprimed 664msec). There was a significant effect of Priming, F(1,22)=39.85, p<O.OOl, but the effect of Decision type did not achieve significance [F(1,22) = 3.1 1, p = 0.091. The Priming x Decision type interaction was significant, F( 1,22) = 12.03, p < 0.005. Post-hoc Tukey tests ( a = 0.05) showed faster RTs to primed than to unprimed familiar faces in the familiarity decision task but not in the expression decision task. The pattern in Experiments 1 to 3 of repetition priming of familiarity decisions but not expression or sex decisions cannot therefore be attributed to slower overall responses in the familiarity decision conditions.

Downloaded by [Bar-Ilan University] at 03:35 19 November 2012

GENERAL DISCUSSION
The main findings of these experiments are as follows:
1. Repetition priming occurs for familiar faces in familiarity decision tasks but not in expression or sex decision tasks. This is not an artefact of the fact that familiarity decisions are typically slower than expression or sex decisions. Simply perceiving, encoding, or responding to a familiar face twice is not sufficient to generate repetition priming. 2. The nature of the initial response to a familiar face is irrelevant to the later manifestation of repetition priming in a familiarity decision task. Thus, deciding that a familiar face is female or that it is smiling causes as much priming when that face is later responded to in a familiarity decision task as deciding that the familiar face i familiar or that it belongs to a politician. s 3. Expression and sex decisions are unaffected by the familiarity or otherwise of the faces being responded to.

These findings, which hold across subjects and across items in all experiments, are all compatible with the theory that repetition priming occurs within the system that stores representations of the appearance of familiar faces. The results pose problems for theories that hold that repetition priming depends on episodic records of previous acts of stimulus encoding (Jacoby, 1983a, 1983b). If those episodic records were memories of

51 0

ELLIS, YOUNG, FLUDE

previous perceptual encodings of seen faces, then the second encounter with a face should always show repetition priming, even when the decision required on the second encounter is a sex or expression judgment. Yet we have consistently failed to observe priming under such conditions. It could alternatively be argued that repetition priming depends on an episodic record of the act of perceiving a stimulus plus the nature of the response made to that stimulus. There was, however, no repetition priming in the expression or sex decision conditions of Experiments 2 and 3, even though exactly the same decisions were being repeated to exactly the same photographs. Repetition priming only occurs in tasks that require that the familiarity and identity of a face be determined. Such tasks include measuring the recognition thresholds for familiar faces, making familiarity or occupation decisions to familiar faces, or naming them. They do not include making sex or expression judgments about the same faces. This is because the systems that respond to sex and expression do not store descriptions of the faces of familiar individuals. Those systems do not show repetition priming, are unaffected by whether a face is familiar or unfamiliar, and can continue to function normally in prosopagnosic patients who have suffered impairment to the system which determines face familiarity. Ellis, Young, Flude, and Hay (1987) showed that repetition priming is sensitive to the degree of similarity between two different photographs of a familiar face when one is used as the priming stimulus and the other as the test stimulus. This graded similarity effect shows that the representations used to determine the familiarity of seen faces are not abstractive logogentype representations, as originally proposed by Hay and Young (1982) and Bruce and Valentine (1989, because units of that sort should be insensitive to changes in the surface form of familiar stimuli (Morton, 1979a).The findings of the present experiments can, however, be reconciled with the graded similarity effect if it is argued that the representations of familiar faces are a set of overlain and overlapping instances in which the most recently encountered instances retain a degree of primacy. This is the type of representation utilized by distributed memory models (McClelland & Rumelhart, 1985). Repetition priming is only found for decisions that require that a face be identified, but any encounter with a familiar face appears to be sufficientto generate repetition priming. Congruence between the responses made to a face at first and second encounter appears irrelevant to the magnitude of repetition priming observed. The implication of this finding is that although the face processing system that responds to familiarity is not normally involved in making sex and expression judgments, representations within that system are activated automatically whenever a familiar face is seen. It is impossible not to recognize a familiar face, whatever decision is explicitly being made regarding that face, and incidental recognition yields full repetition priming. In Fodors (1983) terms, the activation of representations of familiar faces is mandatory and unstoppable.

Downloaded by [Bar-Ilan University] at 03:35 19 November 2012

REPETITION PRIMING AND FACE PROCESSING

51 1

REFERENCES
Bruce, V. (1986). Influences of familiarity on the processing of faces. Perception, 15, 387-397. Bruce, V., Ellis, H. D., Gibling, F., & Young, A. W. (1987). Parallel processing of the sex and familiarity of faces. Canadian Journal o Psychology, 41, 510-520. f Bruce, V. & Valentine, T. (1985). Identity priming in the recognition of familiar faces. British Journal o Psychology, 76, 373-383. f Bruce, V. & Young, A. W. (1986). Understanding face recognition. British Journal of Psychology, 77, 305-327. Bruyer, R., Laterre, C., Seron, X., Feyereisen, P., Strypstein, E., Pierrard, E., & Rectem, D. (1983). A case of prosopagnosia with some preserved covert remembrance of familiar faces. Brain and Cognition, 2, 257-284. Ellis, A. W., Young, A. W., Flude, B. M., & Hay, D. C. (1987). Repetition priming of face recognition. Quarterly Journal of Experimental Psychology, 39A, 193-2 10. Ellis, A. W., Young, A. W., &Hay, D. C. (1987). Modelling the recognition of faces and words. In P. E. Morris (Ed.), Modelling cognition. Chichester: J. Wiley & Sons. Feustel, T. C., Shiffrin, R. M., & Salasoo, A. (1983). Episodic and lexical contributions to the repetition effect in word identification. Journal of Experimental Psychology: General. 112, 309-346. f Fodor, J. (1983). The modularity o mind. Cambridge, MA: MIT Press. Hay, D. C. & Young, A. W. (1982). The human face. In A. W. Ellis (Ed.), Normality and pathology in cognitive functions. London: Academic Press. Humphreys, G. W., Besner, D., & Quinlan, P.T. (1988). Event perception and the word f repetition effect. Journal o Experimental Psychology: General, 117, 5 1 6 7 . f Jacoby, L. L. (1983a). Perceptual enhancement: Persistent effects of an experience. Journal o Experimental Psychology: Learning, Memory and Cognition, 9, 21-38. Jacoby, L. L. (1983b). Remembering the data: Analysing interactive processes in reading. Journal of Verbal Learning and Verbal Behavior, 22,48%508. McClelland, J. L. & Rumelhart, D. E. (1981). An interactive activation model ofcontext effects in letter perception: Part I . An account of basic findings. Psychological Review, 88,375407. McClelland, J. L. & Rumelhart, D. E. (1985). Distributed memory and the representation of f general and specific information. Journal o Experimental Psychology: General. 114, 159188. Morton, J. (1969). Interaction of information in word recognition. Psychological Review, 76, 165-178. Morton, J. (1979a). Facilitation in word recognition: Experiments causing change in the logogen model. In P. A. Kolers, M. Wrolstad, & H. Bouma (Eds.), Processing of visible language. Vol. I . New York: Plenum. Morton, J. (1979b). Word recognition. In J. Morton & J. C. Marshall (Eds.), Psycholinguistics series, Vol. 2: Structures and processes. London: Elek, and Cambridge, MA: MIT Press. Neisser, U. (1954). An experimental distinction between perceptual process and verbal ref sponse. Journal o Experimental Psychology, 47, 339402. Roediger, H. L. & Blaxton, T. A. (1987). Effects of varying modality, surface features, and retention interval on priming in word-fragment completion. Memory and Cognition, 15,379388. Scarborough, D. L., Cortese, C., & Scarborough, H. S. (1977). Frequency and repetition effects in lexical memory. Journal of Experimental Psychology: Human Perception and Performance, 3, 1-17. Tranel, D., Damasio, A. R., & Damasio, H. (1988). Intact recognition of facial expression, gender, and age in patients with impaired recognition of face identity. Neurology, 38, 6% 696.

Downloaded by [Bar-Ilan University] at 03:35 19 November 2012

51 2

ELLIS, YOUNG, FLUDE

Tulving, E., Schacter, D. L., & Stark, H. A. (1982). Priming effects in word-fragment completion are independent of recognition memory. Journal of Experimental Psychology: Learning, Memory and Cognition, 8, 336-342. Young, A. W., McWeeny, K. H., Hay, D C., & Ellis, A. W. (1986a). Access to identity-specific semantic codes from familiar faces. Quarterly Journal of Experimental Psychology, 38A, 271295. Young, A. W., McWeeny, K. H., Hay, D. C., & Ellis, A. W. (1986b). Matching familiar and unfamiliar faces on identity and expression. Psychological Research, 48,6368.
Manuscript received 23 October 1989

Downloaded by [Bar-Ilan University] at 03:35 19 November 2012