... some confinnation of a context effect in the familiarity decision. Key words: Context, face, semantic memory. Mots des : Contexte, visage, memoire semantique.
- - - - - - - - - - - - - - - - - - - - - - - -- - - - --·
C.P.C. : Cahiers de Psychologie Cognitive European Bulletin of Cognitive Psychology 1991, Vol. 11, n"l, 55-71.
EF:F'ECTS OF THE MEANING OF VISUAL CONTEXT ON SEMANTIC PROCESSING OF FAMOUS FACES
Myriam Schweich,1 Anne-Caroline Schreiber,2 Stephane Rousset,2 Raymond Bruyer,1 and Guy Tiberghien2 1. Universite de Louvain (UCL), Departement de Psychologie Unite de Neuropsychologie Cognitive (NECO) 20 voie du Roman Pays, 1348 Louvain-la-Neuve, Belgium 2. Universite de Grenoble, Laboratoire de Psychologie Experimentale 1251 Avenue Centrale, BP 47 X, 38040 Grenoble Cedex, France
Abstract Two experiments were conducted to test the role and the importance of visual context on the processing of famous faces. The semantic congruence between the identity of the faces and the meaning of the situation induced by the visually associated background was manipulated: the context could be neutra~ congruent, incongruent or with no special meaning (a landscape). In the first experiment, a familiarity decision, a semantic decision (about the occupation), and a naming task were administered to three groups of subjects. ·· Our results are broadly consistent with the predictions derived from Bruce and Young's (1986) functional model of face processing. The classic hierarchy between the three tasks was found in terms of speed of response. A semantic contextual effect arose from 'the meaningful visual background: an inhibitory effect of contextual incongruence was obtained on the three tasks. Bruce and Young's model predicts context effects on access to identity-specific-semantic codes and on name codes, but not on access to face recognition units. In the second experiment, there was some confinnation of a context effect in the familiarity decision.
Key words: Context, face, semantic memory. Mots des : Contexte, visage, memoire semantique.
"Tous droits d.e_QLQpriete intellectuelle reserves . Reproduction representation interdites sans autorisation (code de Ia propriete in tel led~'-·
56
M. Schweich, A.-C. Schreiber, S. Rousset, R. Bruyer, & G. Tiberghien
INTRODUCTION In the literature on face memory, a growing body of studies has examined the effects of context. The term "context'' has been operationalized in various ways (for recent reviews, see Davies, 1986, 1988). Context has been defined as the accompanied face when faces were studied in pairs (Bower & Karlin, 1974; Watkins, Ho, & Tulving, 1976; Winograd & Rivers-Bulkeley, 1977), as clothing (Brutsche, Cisse, Deleglise, Finet, Sonnet, & Tiberghien, 1981; Cuttler & Penrod, 1989), as encoding instructions (Baddeley & Woodhead, 1982; Bower & Karlin, 1974; Daw & Parkin, 1981; Winograd, 1978), as the background in which the face is presented (Davies & Milne, 1982; Daw & Parkin, 1981; Klee, Leseaux, Malai, & Tiberghien, 1982; Memon & Bruce, 1983; Parkin & Hayward, 1983; Peds & Tiberghien, 1984; Thomson, Robertson, & Vogt, 1982), and also as sentences related to the face about a characteristic of the person, such as the occupation (Baddeley & Woodhead, 1982; Kerr & Winograd, 1982; W~tkins et al., 1976). Thus, the term "context" has been given a number of different meanings. The fact that the context in which faces are presented can influence the ability of subjects to recognize or recall the same face has been demonstrated in many empirical studies, but these effects are not invariable or always of the same strength. They depend, for example, on face familiarity (Thomson et al., 1982) or on the type of context (Davies, 1988). A first step, not restricted to faces, in the understanding of this effect is represented by the debate between the network theory of Anderson and Bower (1972), which represents events by nodes in a semantic network, and the "encoding specificity hypothesis" of Tulving and Thomson (1973), in which every encounter with a stimulus creates a unique encoding or episode. In order to explain the fluctuations of context effects, Baddeley and Woodhead (1982) introduced the distinction between interactive and independent contexts, which takes into account the relationship between the context and the stimulus. An independent context has an arbitrary relationship with the stimulus. The change or deletion of this context will have no influence on recognition performance. An interactive context, conversely, will influence the way in which the subject encodes the stimulus since such a context has a significant relationship with the stimulus. Memon and Bruce (1985) proposed a different approach to explain context effects in face recognition in the theoretical framework of Bruce and Young's cognitive model (1986). According to these authors, recognition performance is dependent on the matching between the codes formed during learning and the codes available at the retrieval stage. Their model maintains that a photograph of a face will lead to the generation of a pictorial code. When a subject is presented the same picture of an unfamiliar face, his or her performance will
"Tous droits de propriete intellectuelle reserves. Reproduction reoresentation interdites sans autorisation (code de Ia propriete intellectuelle)".
Visual context and semantic processing offaces
57
depend on remembering pictorial details. Structural codes are used to recognize different views of faces. These two types of code are representations of the · visual information present in a face. Bruce and Young described two kinds of semantic information that can be derived from a face: the visually-derived semantic codes (VDSC) and the identity-specific semantic codes (ISSC). The VDSCs can be derived directly from the surface form of the face, while the ISSCs are the representation of who the person is (occupation, etc.) and applies to familiar faces only. For familiar faces only, the authors then postulated face recognition units (FRU), ISSCs, and name codes. Recent activation of these pre-existing representations can be used to mediate the episodic recognition processing of familiar faces. Within this framework, an explanation of context effects in episodic recognition tasks can be postulated: when an unfamiliar face is presented for recognition in a new context, the VDSCs established at inspection may not match the codes established in the recognition stage, and the face may remain unrecognized. In the same way, a familiar face should not be affected by changes in the background context because, in this case, the FRUs and the ISSCs are available in memory whatever the background. We may thus predict that the only way context could affect recognition memory for familiar faces is when a change in the ISSCs is introduced. Bruce and Young's (1986) model holds that familiar and unknown faces are not processed in the same way. The representation of these different classes of faces in memory is also probably different: familiar faces have a pre-existing structural representation contained in the FRU and a more or less rich semantic representation (ISSC), which is not the case for unknown faces. This theoretical model postulates a feedback from the semantic level to FRUs that could explain the effects of context on the familiarity decision. According to Tiberghien (1986), not only do variations of context act on identification of persons, access to semantic information, or naming, they can also modify the feeling of familiarity. This can be the case when an inappropriate meaningful context leads to doubt about the familiarity of a well-known face or when an unfamiliar face is mistaken for a familiar one because of a specific context (in this respect, seethe daily-life errors recorded by Young, Hay, & Ellis, 1985).
EXPERIMENT 1 The purpose of our experiments concerns the effects of the visual context associated with a face at different stages of the processing. We manipulated the semantic relation between face identity and the information evoked by its visual context. Three tasks were used: a familiarity decision, a semantic decision (about occupation), and a naming task. It was expected that the context congru-
"Tous droits de propriete intellectuelle reserves. Reproduction representation interdites sans autorisation (code de Ia propriete intellectuelle)".
- -- - - -·-·----·-- --- - -- --- ··- ---- - -·-· ·
- -·
58
M. Schweich, A.-C. Schreiber, S. Rousset, R. Bruyer, & G. Tiberghien
ence or incongruence would have no effect on familiarity decisions, because this decision is made at an early stage of the processing (access to FRUs) without necessarily reaching the ISSC level. As a matter of fact, in Bruce and Young's (1986) model, the context characteristics should not act directly on the FRU level. Contextual information could have an effect on FRUs as feedback from the ISSCs to the FRUs, some contextual elements activating the semantic nodes, which in turn would activate the FRUs. This feedback process would not be evoked in a familiarity decision, if one assumes that familiarity decision took place at the FRU level (Young & Ellis, 1989). It would be unreasonable to assume that feedback induced by the semantic processing of context would be faster than the simple activation of the FRU except the case in which context is pre-existent, without totally rejecting the sequentiality and the heuristic interest of Bruce and Young's model. Thus, according to this model, the context should not induce an effect of congruence on the familiarity decision. In the semantic decision, a facilitative effect of congruent context should occur because we are testing ISSCs. In Bruce and Young's (1986) model, the context effects are essentially explained by the close relationship maintained between the ISSCs and the contextual information. A naming task should also be sensitive to the same effects as the occupation decision. Indeed, this task requires access to ISSCs and the model postulates a hierarchy between the ISSCs, and subsequently, the name codes. Four types of context were used in function of the meaning of the relation with the identity of the faces. Especially for famous, well-known faces, the context can be congruent (providing semantic congruent information with the ISSCs), or not (providing semantic information that does not correspond to the associated face). In view of the distinction introduced by Baddeley and Woodhead (1982), these two contexts are supposed to be interactive with the face identity. The context can also have no relationship with the identity of the face, like a landscape without any special meaning (control condition with regards to the congruent and incongruent contexts, with the same visual complexity but less semantic complexity) or it can be a non-figurative context and provide no information at all about the face (neutral context). These two latter types of context are assumed to be independent of the face identity. We expected that a congruent context would facilitate the semantic decision and naming, and an incongruent one would interfere. Familiarity decision should remain free of context effects. Although the landscape context has the same visual complexity as congruent and incongruent contexts, it should not bring information about the identity of the face (base-line condition). This condition should permit us to evaluate the base-line with regards to the identity-specific semantic codes. The neutral context was introduced as a control in the experiment because of the visual and semantic complexity of the context.
interdites sans autorisatior]J!;Q_de de I
~
+
.0
E :J
Famous Unknown
c:: c::
cu