Journal of Personality and Social Psychology Copyright 1981 by the American Psychological Association. Inc.

1981, Vol. 40, No. 3, 441-452 0022-3514/81 /4003-0441 J00.75

Person Categories and Social Perception:

Testing Some Boundaries of the Processing
Effects of Prior Knowledge
Claudia E. Cohen
Rutgers—The State University

Current assumptions about the processing role of prior social knowledge in social
perception may need modification when applied to situations that are richer and
more complex than typical research paradigms. Two experiments investigated
whether stereotypic knowledge would influence social perception in a more re-
alistic setting. In Experiment 1, subjects watched a videotape of a target woman
identified either as a waitress or a librarian. Subjects more accurately rememA
bered features of the woman that were consistent with their prototype of a /
waitress (librarian) than features that were inconsistent. The prototype-consis-
tency effect did not interact with the delay time before recognition memory was
assessed. In Experiment 2, subjects learned the occupational information either
before or after watching the tape. The prototype-consistency effect from Exper-
iment 1 was replicated. In addition, knowing the target's occupation while watch-
ing her led to increased accuracy for both consistent and inconsistent information.
The probable role of both encoding and retrieval processes in contributing to this
effect is noted. Perceivers' stereotypic prior knowledge influenced their memory /
of a target person's behavior even in a realistic person-perception situation. Con-
ditions that favor the memorability of consistent versus inconsistent information
are discussed.

Recent work in person perception has em- age, or retrieval of social information (cf.
phasized the influence of prior social knowl- Taylor & Crocker, 1980). Acknowledging
edge, structured and stored in the perceiver's the enormous richness of the social environ-
"head," on the processing of novel infor- ment, most recent theorizing about social
mation about another person (Cantor & cognition has argued that a perceiver's prior
Mischel, 1977, 1979; Hastie & Kumar, knowledge serves to structure and reduce the
1979; Taylor, Fiske, Etcoff, & Ruderman, flow of incoming information, preventing
1978). In this view, the perceiver is an active cognitive chaos. Yet supporting research has
processor whose social knowledge in the been conducted in paradigms that greatly
form of "schemas," "prototypes," or "cate- simplify the richness and complexity of the
gories" may affect his or her encoding, stor- person-perception situation. Though this
simplification was probably appropriate in
the initial stages of investigation, one may
The first experiment described here was part of a now ask how conclusions based on prior re-
thesis submitted to the Department of Psychology,
University of California, San Diego, in partial fulfill- search apply beyond the controlled settings
ment of the requirements for the doctoral degree. Por- used in those studies.
tions of this study were reported at the 85th annual APA This article has two aims. First, some lim-
Convention, San Francisco, 1977. The research was fa- iting features of current person-perception
cilitated by conversations with Ebbe Ebbesen and Don- paradigms will be considered, and a study
ald Norman and by the assistance of Cathi Davis, Karen
Bronikowski, and Lucille Traina. Thanks are due Dick designed to circumvent these limitations will
Ashmore, Richard Bowers, Nancy Cantor, and two be presented. Second, two specific processing
anonymous reviewers for their very helpful comments issues will be considered: the relative mem-
on an earlier draft of the manuscript. orability of information that is consistent or
Requests for reprints should be sent to Claudia E.
Cohen, Department of Psychology, Douglass College, inconsistent with a social category and the
Rutgers—The State University, New Brunswick, New role of encoding and retrieval in category-
Jersey 08903. based processing. Research designed to cre-
441
442 CLAUDIA E. COHEN

ate a rich, realistic person-perception setting tures should be considered in more depth.
should contribute to the accumulating (and These are the mode of information presen-
sometimes conflicting) evidence about these tation, the complexity of the person infor-
processing issues. mation, and the relationship between the
form of the information presented and the
Limitations of Traditional Person- form of the response required of subjects.
Perception Paradigms First, when information is presented via a
written description, certain types of visual
Consider the typical paradigm used in information are obviously omitted (e.g., non-
person-perception research. Subjects are verbal cues, social setting, etc.) Evidence
given a written set of trait adjectives (or brief suggests that memory for visually presented
behavioral sequences) attributed to a ficti- stimuli, especially people's faces, tends to be
tious target person and then respond to judg- more accurate than for comparable verbally
ment or memory measures, also in the form presented information (Bahrick, Bahrick,
of trait ratings (or brief behavioral descrip- & Wittlinger, 1975; Goldstein & Chance,
tions). Compare this paradigm with real-life 1971; Shepard, 1967). Since perceivers in
person-perception situations encountered face-to-face interactions can rely on visual
outside of the laboratory. The perceiver, ob- information rather than solely on abstracted
serving the target person's behavior as it verbal descriptions to accurately fix an ac-
unfolds, focuses on selected features as he tor's behavior in memory, they may be less
or she structures the incoming information.' influenced by their prior stereotypic knowl-
His or her memory for that information may edge. This suggests that perceivers in real-
then become the basis for the perceiver's own life social situations may more accurately
judgments, attributions, predictions, and be- remember actors' behavior than current so-
haviors. cial perception research suggests.
Given that the typical paradigm clearly A second feature concerns the tendency
simplifies the perceiver's task, what is the for the information that is presented to the
impact on our conceptual analysis of the perceiver to be relatively simple and one-di-
person-perception process? Current ap- mensional. Traits and behaviors that are as-
proaches assume that the perceiver has avail- cribed to a target person usually relate to
able two sources of input in a particular sit- only one dimension or aspect of his or her
uation: (a) prior, structured knowledge about personality (e.g., how likable the person is),
people and the social world and (b) the spe- corresponding to the category scheme the
cific behavior of the target person. The rel- experimenter is investigating. Outside of the
ative importance of the perceiver's prior laboratory, of course, many features may be
knowledge has been stressed by many in- used by a perceiver as a basis for categoriz-
vestigators (cf. Taylor & Crocker, 1980). ing a target person: demographic character-
Yet other evidence suggests that target in- istics (e.g., sex, race), personality charac-
formation, especially when it is "concrete, teristics (e.g., introversion-extra version),
vivid and salient," may override more ab- occupation, and so on (Ashmore & Del
stract forms of knowledge (Nisbett & Bor- Boca, 1979; Brigham, 1971; Cantor & Mis-
gida, 1975, p. 943). That is, information ob- chel, 1979; Hamilton, 1979; Rothbart, Fu-
served in face-to-face interaction may take lero, Jensen, Howard, & Birrell, 1978). Fre-
precedence over a perceiver's abstract ste- quently, these different category schemes
reotypic expectations. This issue—when and may contain conflicting information about
if concrete target information will override the target person: for example his boyish
prior knowledge—cannot be evaluated ade- appearance suggests that he is young and
quately with results from research in which naive, whereas his colorful vocabulary re-
perceivers did not actually observe behavior.
A paradigm that includes the on-line obser-
vation of behavior seems clearly warranted. 1
It is not assumed that the perceiver consciously se-
To illustrate the conceptual boundaries lects features of the target person or that he is aware
imposed by a simplified paradigm, three fea- of the features he has either focused on or ignored.
 PERSON CATEGORIES AND SOCIAL PERCEPTION 443

veals that he is "streetwise." If conflicting To reasonably apply previous research re-

category schemes cancel out one another sults to real-world social perception, two ad-
(Carlston, 1980), the role of implicit prior ditional issues need to be addressed. Con-
information may be minimized. This prob- sider an example: A student with a well-
lem of multiple, conflicting category schemes defined prototype of a typical professor
is not currently reflected within the typical (Cantor & Mischel, 1979; Cohen, Note 1)
person-perception paradigm. meets a new member of the faculty to discuss
A final, more subtle restriction of the tra- a course. The professor is both consistent and
ditional paradigm is that researchers have inconsistent with the student's image; for
frequently maintained a high degree of sim- example, her office is filled predictably with
ilarity between the form of the stimulus in- professional journals and incongruously with
formation about the target person and the stacks of racing forms. When the student
form of the response that the perceiver is remembers the interaction at some later
required to make (i.e., the dependent vari- date, will his view of her office contain the
able). As noted, both stimulus and response journals, the racing forms, or both? Assume
measures have typically (though by no means that in this instance the consistent infor-
always) taken the form of written trait terms mation is remembered better. This outcome
(or behaviors) describing the target person.2 could be due to encoding selectivity (failure
This procedure substantially limits the flex- to encode the stack of racing forms), mem-
ibility that perceivers have outside of the lab- ory retrieval (prototype-inconsistent infor-
oratory to structure or "unitize" a target mation may be difficult to locate), or both.
person's stream of behavior in different In the latter case, with an effective strategy,
ways. Recent evidence suggests that the the student may be able to "find" the in-
same information may be structured differ- consistent information; in the former case,
ently by two perceivers (e.g., as a result of the information is obviously nonretrievable.
their observational goals; Cohen & Ebbesen,
1979; Jeffery & Mischel, 1979), influencing Processing Issues
the subsequent availability of the informa-
tion. When the informational units at the Consistent and inconsistent information.
points of encoding and responding are po- There is evidence in the social cognition lit-
tentially more dissimilar (e.g., the perceiver erature for superior memorability of both
observes a videotape of a target person and category-consistent (e.g., Rothbart, Evans,
then makes trait ratings about him), the im- & Fulero, 1979) and category-inconsistent
pact of prior knowledge may be minimized information (e.g., Hastie & Kumar, 1979).
as the person information is "translated" Previous research suggests several factors
from one structured form to another. This that may promote the memorability of con-
is consistent with the notion of encoding sistent and inconsistent information. First,
specificity in which retrieval of information consistent information about a target person
from memory depends on the initial form in may be favored in a paradigm that contains
which it was encoded (Tulving & Thomson, a diverse array of information rather than
1973). one that is limited only to category-consis-
In sum, several features of previous social tent and inconsistent information (Rothbart
perception research—mode and complexity et al., 1979). Prior knowledge may reduce
of person stimuli, similarity between stim- the information load on the perceiver by fa-
ulus and response forms—may restrict the voring consistent information.
general applicability of its results. Due to Second, the processing advantage of in-
these restrictions, conclusions drawn from consistent information may be associated
this body of research may need to be mod- with a set-size effect. Hastie and Kumar
ified when person-perception settings extend (1979) found that inconsistent characteris-
beyond the boundaries of the typical para- 2
Although previous studies have occasionally used
digm. Research will be presented here in realistic videotaped stimuli (e.g, Zadny & Gerard,
which the person-perception setting was de- 1974), the conceptual limitations of paper-and-pencil
signed to eliminate these typical limitations. stimuli have not been addressed in prior research.
444 CLAUDIA E. COHEN

tics of a target person are less likely to be tion, eliminating the role of encoding (e.g.,
selectively remembered as the proportion of Rothbart et al., 1979; Zadny & Gerard,
category-consistent and category-inconsis- 1974). Although these data suggest that re-
tent features of the target person becomes trieval-only effects are minimal, more evi-
approximately equal. Finally, manipulations dence is needed to clarify a somewhat mixed
of the perceiver's observational goal have pattern of results. Previous studies cannot
produced contradictory results about consis- be conclusively compared because of varying
tency and memorability. When the perceiver stimuli and diverse dependent measures.
is instructed to form an impression of the Because of the intentionally realistic par-
target person, some evidence suggests that adigm used in the current research, the pat-
he or she will better remember information tern of results might differ from previous
that is incongruent with an original category studies on a related issue. In prior research,
assignment (Hastie, 1980). However, other category-based processing effects, whether
research indicates that an impression-form- caused by encoding, retrieval, or both, were
ing goal encourages perceivers to reduce the apparent immediately after subjects viewed
complexity and thus overlook incongruent the target person (e.g., Cantor & Mischel,
information (Cohen, in press). Although 1977). In the current study, it was expected
there is no clear prediction from these stud- that these processing effects might not ap-
ies, the observer's goal may play an impor- pear until somewhat later, when perceivers
tant role in the consistency and memorabil- had forgotten details of the lifelike stimulus.
ity of information. The first experiment presented here inves-
In this current research, it was predicted tigated this issue, simultaneously collecting
that category-consistent information would indirect evidence about the relative effects
be more memorable. The paradigm incor- of encoding and retrieval. If a delay period
porated conditions that should have pro- of several days or more was required before
moted memorability of consistent informa- category-based effects were evident, then
tion: The stimulus materials were complex retrieval processes are implicated. Con-
and realistic, and the target person was versely, any effects of encoding selectivity
"equally" consistent and inconsistent with should have been apparent immediately af-
a particular category prototype. In addition, ter the target person was observed.
the research assumes a view of social cog-
nition in which knowledge about category Empirical Investigation
members is organized around the category
prototype (cf. Cantor & Mischel, 1977; Perceivers watched a videotaped vignette
Rosch, 1975; Cohen, Note 1). Theoretically, of a woman and her husband engaged in a
processing should favor category features realistic interaction. The woman's occupa-
that most closely resemble this prototype. tion was selected as the category that would
Encoding and retrieval: When do the ef- be manipulated, and the study was designed
fects of category-based processing occur"] to determine whether occupation knowledge
The relative effects of encoding and retrieval would influence perceivers' memory for life-
processes in person perception are difficult style, appearance, and other nonpersonality
to separate. Several studies have found evi- information. First, extensive data were col-
dence for category-based processing when lected to assess subjects' prototypes of the
subjects were given a category label for a target occupations, waitress and librarian.
stimulus person prior to receiving other trait Based on these data, a videotape that con-
or behavioral information about them (cf. tained half librarian and half waitress fea-
Taylor & Crocker, 1980). Viewed alone, this tures was carefully constructed. To create
result could be attributable to encoding or realistic complexity, the videotape also con-
retrieval processes or their combination. tained dialogue and behavior that were not
However, category-based effects were much part of subjects' prototype descriptions (e.g.,
less powerful when subjects received the cat- they talked about freeway driving, discussed
egory label after the other person informa- his parents, etc.) The videotape was shown
 PERSON CATEGORIES AND SOCIAL PERCEPTION 445

to a second group of subjects, instructed to dren), and her occupation (e.g., "works full time in a
form an impression of the target woman's coffee shop"). They were instructed to imagine the
woman vividly and then to describe her physical ap-
personality. These subjects were casually pearance, home environment, and life-style preferences
told the woman's occupation: either a wait- (e.g., in music, sports, books, food, etc.). After com-
ress or a librarian. After viewing the tape, pleting and returning the first description, subjects were
subjects were asked questions about details given information about a second woman, differing only
in occupation and initials (e.g., "E. N., a librarian em-
of the woman's physical appearance, behav- ployed in a downtown public library"). Subjects then
ior, and life-style. Of course, these questions described the second stimulus person, responding to the
were based on the "waitresslike" and "li- same descriptive categories as with the first woman (e.g.,
brarianlike" features of the tape. her appearance, etc.). Order of the occupations was
As noted, a secondary aim of Experiment counterbalanced across subjects.
1 was to determine whether category-based 10Selection of items for ratings. The 10 waitress and
librarian descriptions were summarized as follows.
processing would affect memory immedi- Two trained raters independently identified all distinct
ately after the target woman's behavior was features that appeared in the pool of descriptions and
observed or whether, in this richer context, tabulated (for each stereotype separately) the number
some passage of time would be required. To of subjects who mentioned the attribute. The two raters'
estimated frequencies proved highly reliable (across the
test this, the delay period between subjects' 245 features, r = .91, p < .001).
viewing of the videotape and responding to From these descriptions, a pool of 82 features from
the memory questions was manipulated. As 21 different categories was selected. Features were cho-
the delay increased, it was assumed that the sen on the basis of both absolute frequency (i.e., the
details of the tape would fade and subjects absolute number of times they were mentioned) and
relative frequency (i.e., the items chosen were selectively
would rely more heavily on categorical pro- more frequent for one of the prototypes, waitress or li-
cessing. These subjects should make more brarian). One to five features were selected from each
errors by incorrectly retrieving prototypical of the 21 categories.
(or category-consistent) features than sub- Ratings of prototypes. Twenty-seven subjects (11
jects who were questioned immediately. their male and 16 female) received experimental credit for
participation. They were asked to imagine their
However, if subjects who were questioned "clearest example" of a waitress (or librarian) and,
immediately after viewing a realistic vi- keeping this image in mind, to rate the relevance to it
gnette were selectively accurate for stereo- of each of a series of features. For instance, they were
typic information, the potency of category- asked to rate (on a 7-point scale ranging from "not at
all likely" to "extremely likely") how likely the target
based processing would be highlighted. person was to "have blonde hair." When several features
from the same category were included, subjects were
Experiment 1 not constrained to choose one item but rather rated each
one. For example, they rated her preference separately
There were two phases to this study: the for classical music, pop, jazz, "easy listening," and
"music of the 30s and 40s." This method avoided cre-
assessment of prototypes for the two target ating an artificial contrast between the two stereotypes.
occupations and the videotape memory ex- Each subject rated both prototypes on all 82 features,
periment. Because no relevant descriptions and prototypes were counterbalanced for order across
of prototypical life-style or appearance were subjects.
available, this assessment was an important Selection of features for videotapes. Eighteen pairs
of features from 12 different categories were chosen for
part of the study. the videotape experiment. Each feature pair was drawn
from the same category and consisted of one feature
Method that was rated highly likely for the waitress prototype
and one feature that was rated highly likely for the
Prototype Assessment librarian prototype. For example, one appearance fea-
ture pair matched the librarian feature "wears glasses"
Open-ended prototype description. To identify im- with the waitress feature "does not wear glasses." The
portant characteristics associated with the prototypical feature pairs were split so that one appeared in each of
waitress (librarian), open-ended descriptions were col- two videotapes, Tapes A and B. Table 1 presents the
lected. Five male and five female undergraduate subjects complete set of features.
from the same population as that of the subsequent Each tape contained half of the librarian features and
phases of the study were given minimal information half of the waitress features: Tape A contained Librar-
about a female target person: that is, her initials, her ian Features 1-9 and Waitress Features 10-18; Tape
marital and family status (i.e., married with two chil- B contained Librarian Features 10-18 and Waitress
446 CLAUDIA E. COHEN

Table 1 fered only on the specified 18 pairs of features. These

Manipulated Videotape Content: Prototype features were slipped subtly into the tapes (e.g., the two
Features in Tapes A and B tapes differed on whether there was artwork on the wall)
or were integrated into the conversation (e.g., she men-
Tape A Tape B tioned playing either the piano or the guitar). Target
features were clearly associated with the woman rather
Librarian features than with her husband. Throughout the videotape, the
woman and her husband carried on a lively conversation
Roast beef Salad while eating dinner and later while opening her birthday
Plays piano Wine presents. The actors were married in real life so they
Fresh flowers Formal table setting portrayed the relationship quite naturally; and the vid-
Wears glasses Traveled in Europe eotapes were recorded in a home to lend further realism.
Bookshelves No television Procedure. Three subjects at a time were individu-
Angel food birthday cake Spent day reading ally seated in separate booths, and a large 19 in. (48.2
Golf clubs Artwork cm) Sony videotape monitor and deck were located on
Classical music Receives bestseller as gift a table approximately 3 ft. (.91 m) in front of the sub-
Nonaffectionate with Receives history book as jects. The experimenter explained that she was inves-
husband gift tigating how people form impressions of other people.
Subjects were instructed to watch the target person and
Waitress features form an impression, as if it were a real-life situation.
Hamburgers At this point the occupation manipulation was casually
No salad introduced: The experimenter mentioned either that the
Beer Plays guitar
target woman worked as a waitress in a local coffee shop
Informal table setting No fresh flowers or as a city librarian. All subjects were told that the
Has not traveled in Does not wear glasses
No book shelves couple had consented to be videotaped in their own
Europe home.
Television Chocolate birthday cake
Spent day working Bowling ball The experimenter told subjects that after viewing the
No artwork Pop music videotape, they would be asked questions about the per-
Receives nightgown as Affectionate with sonality traits of the target woman. Subjects then
gift husband watched one of the two videotapes—either Tape A or
Receives romantic novel Tape B-—depending on a random assignment. Again,
as gift the tape condition was orthogonal to the occupation con-
dition (e.g., half the librarian-set subjects watched Tape
A and half watched Tape B). Only 60-sec lead-in se-
quences differed for the waitress- and librarian-set con-
Features 1-9. Thus, the two tapes were equally "wait- ditions. In each of these segments, the woman entered,
resslike" and "librarianlike". wearily sat down, and briefly described her day at work
(e.g., the waitress complained about serving endless cups
of coffee). These lead-in segments were meant to
Observation of Videotape Vignette strengthen her credibility as a member of the appro-
priate occupation. No further mention of the woman's
Overview. As part of an impression-formation study, occupation was made.
96 undergraduates (48 males and 48 females) at a state
university in California viewed a videotape of a woman After viewing the videotape, subjects gave their
and her husband. Half of the subjects were told the impression of the woman, using a 37-item trait ques-
target woman worked as a waitress (waitress set), and tionnaire. Subjects in the two delay conditions were then
the other half were told she worked as a librarian (li- asked to return—either 4 or 7 days later. They were
brarian set). All subjects answered recognition memory erroneously told they would watch a videotape of a dif-
questions about the woman's appearance and behavior. ferent person at that time, to prevent rehearsal of the
Crossed with occupation set was the length of the delay current videotape in anticipation of further questioning.
between subjects' viewing of the tape and answering Subjects in the immediate (no-delay) condition were
questions about it: no delay, 4 days, or 7 days. Two advised that they would next be asked some slightly
videotapes were used to add generality to the results; different questions about the target woman. After sub-
the specific videotape that a subject watched was a third jects completed the memory question booklet, the pur-
between-subjects factor (Tape A, Tape B). Eight sub- pose of the experiment was explained.
jects were randomly assigned to each of the 12 cells When delay subjects returned for a second session,
(2 X 3 X 2). In addition, within-subjects, memory items they were immediately presented with memory booklets.
were drawn half from the prototypical waitress and half The experimenter explained that first she would ask
from the prototypical librarian descriptions (item type). them a few more questions about the woman whom they
Videotape stimuli. Two 15-min. videotapes (Tapes had viewed last time and then restated the woman's
A and B) were constructed, using the same professional occupation, as if to verify which videotape they had
actress and actor. Each videotape, complete with sound, watched. The rest of the procedure was identical to the
showed the actor and actress first having dinner and one for the immediate subjects.
then having an informal birthday celebration. Tape A Dependent measures. The memory measure con-
and Tape B, constructed to be virtually identical, dif- sisted of 18 two-alternative forced-choice questions
 PERSON CATEGORIES AND SOCIAL PERCEPTION 447

about the videotape, based on the 18 pairs of features The pattern of accuracy results across de-
mentioned above. One response alternative was always
correct (i.e., it had appeared in the tape), and the other
lay intervals is presented in Table 2. Con-
was always incorrect. Also, as noted above, one alter- trary to prediction, the selective accuracy
native on each question was a waitress feature (e.g., ate effect was approximately constant across the
hamburger) and one was a librarian feature (e.g., ate three delay intervals; the interaction with
roast beef). Thus, if a subject answered all of the ques- length of delay was not significant. Further-
tions correctly, half of his or her responses would be
waitress features and half would be librarian features. more, this selective accuracy for prototype-
consistent features was significant, even in
the immediate delay condition, F(l, 84) =
Results and Discussion 8.46, p < .005.3
Two groups of subjects watched a video- The results of this study demonstrated
taped vignette. One group was told that the that mere knowledge of an individual's oc-
target woman was a waitress, and the other cupation influenced perceivers' memory for
group was told she was a librarian. Did the person's life-style and behavior, even in
knowledge of her occupation affect how sub- a rich and complex person-perception set-
jects remembered the details of the vignette? ting. Perceivers were selectively more ac-
Or did the richness and salience of the vid- curate in remembering those characteristics
eotape information override occupation- that fit their prototype than those features
based categorical processing? Analyses were that were inconsistent with the target per-
performed to answer these questions. son's occupation. The processing advantage
for prototype-consistent information was ap-
parent immediately as well as several days
Influence of the Occupational Prototype
later. This result suggests that category-
on Memory Accuracy Results
based processing may have a strong impact
If knowledge of the woman's occupation when perceivers report on even very recent
facilitated processing of features that were social interactions.
prototype consistent, waitress-set subjects In addition, contrary to expectation, a
should have recognized waitress features longer delay interval did not strengthen the
more accurately, whereas the reverse pattern effect of categorical processing. This pattern
should have occurred for librarian-set sub- of results, as noted by Hastie (1980), sug-
jects. An analyses of variance was performed
3
on the recognition memory data. As hypoth- Because the dependent measure was a two-alter-
esized, subjects were more accurate on pro- native forced choice, the data discussed here as accuracy
data may be viewed alternatively as error data. When-
totype-consistent items than they were ever subjects incorrectly chose a feature that was con-
on prototype-inconsistent items; percentage sistent with their prototype but not presented in the
correct out of nine items = 78% and 71%, videotape, they simultaneously decreased their accuracy
respectively, F(l, 84) = 17.08, p < .0001. for actually presented, inconsistent features and made
a prototype-consistent error. Thus, the forced-choice
measure may reflect a possible combination of proto-
Effects of the Prototype on Memory type-consistent accuracy and prototype-inconsistent er-
Accuracy Across Delay Intervals rors.

It was assumed that subjects' memory for

the tape would be less accurate as the time Table 2
between viewing the tape and answering the Memory Accuracy as a Function of Prototype
memory questions increased. The expected Consistency and Delay Interval
decrease in memory accuracy did occur; per- Length of delay interval
centage correct for the immediate condi- Prototype
tion = 83%; for the 4-day delay = 71%; for features Immediate 4 days 7 days
the 7-day delay = 70%; F(2, 84) = 21.46,
p < .0001. Thus, several days after viewing Consistent .88 .73 .73
Inconsistent .78 .68 .66
the videotape, subjects were less able to rec-
ognize life-style features that they had ac- Note, Average percentage correct out of nine recognition
tually seen. items.
448 CLAUDIA E. COHEN

gests that encoding or retrieval processes last scene was eliminated.4 In addition, the lead-in se-
accounted for the effect rather than differ- quence restating the woman's profession was eliminated,
since half of the subjects would not learn her profession
ential decay rates of consistent and incon- until after viewing the tape.
sistent information. If category-consistent
effects had not appeared in the immediate
condition, the role of selective encoding Procedure
could have been eliminated. However, given The general procedure was identical to that of Ex-
the current pattern of data, the relative roles periment 1. Subjects in the "set after" condition were
of encoding and retrieval cannot be sepa- given the same impression formation instructions before
viewing, but no mention was made of the target woman's
rated. occupation until after they had viewed the tape. As be-
To separate these processes, a condition fore, the experimenter casually mentioned the woman's
is needed where perceivers learn the relevant occupation (i.e., waitress, librarian) without stressing
categorical information after viewing the this information. After viewing the tape (and receiving
target's behavior, eliminating the possible the occupation information), all subjects filled out the
same trait questionnaire and recognition memory items
role of selective encoding. Although not a as in Experiment 1. Because the tape was shortened,
definitive test of the relative roles of encod- however, some memory items were necessarily elimi-
ing and retrieval, a second videotape study nated.
was conducted to pursue this issue. If en-
coding of prototype-consistent features con- Results and Discussion
tributed to the effect, subjects who knew the
target's occupation while viewing the tape Replicating the results from Experiment
should be more selectively accurate than 1, subjects were more accurate for proto-
subjects who only learned her occupation type-consistent features than for prototype-
after viewing the tapes (e.g., Rothbart et al., inconsistent features: 71% and 62%, respec-
1979; Zadny & Gerard, 1974). Alterna- tively, F(l, 50) = 7.17, p < .01. If encoding
tively, if retrieval processes were more im- processes contributed to this effect, subjects
portant, the timing of the occupation infor- who knew the target woman's occupation
mation should have had no effect and before the videotape should have been more
selective accuracy should be equal in both selectively accurate than subjects who re-
conditions. ceived this information after the videotape,
producing an interaction between set timing
and selective accuracy. However, the timing
Experiment 2 effect did not interact with selective accu-
racy, as shown in Table 3 (F < 1, ns). Know-
Method ing the woman's occupation prior to watch-
ing the vignette facilitated subjects' accuracy
Overview equivalently for both prototype-consistent
and prototype-inconsistent features.
Much of Experiment 2 was identical to Experiment
1. Subjects (56 undergraduates: 32 female and 24 male) Despite the nonsignificant interaction, the
watched the same videotape of the target woman and overall level of accuracy did differ as a func-
her husband. Half were told that she was a waitress, tion of set timing. Subjects who knew the
and the other half were told that she was a librarian. target woman's occupation while observing
In addition, set timing was crossed with occupational the videotape were more accurate across
set: Half were told about her occupation prior to viewing
the tape, as in Experiment 1, and the other half were item type: before = 70%; after = 62%; F(l,
not told until just after viewing the tape. All subjects 50) = 5.26, p < .03. Thus, knowing the tar-
answered the recognition memory questions during the get woman's occupation while viewing the
same session in which they viewed the tape. videotape enabled subjects to more accu-
4
The pattern of results was essentially the same for
Videotape Stimuli Tapes A and B. Since subjects were slightly less accurate
on Tape B, it was chosen to allow more variation in
A subset of the videotape stimuli from Experiment selective accuracy. In addition, the last scene in this tape
1 was used for comparison with the previous results. To was eliminated to avoid a ceiling effect, because those
simplify the procedure, only Tape B was used, and the items were particularly well remembered.
 PERSON CATEGORIES AND SOCIAL PERCEPTION 449

Table 3 might have used to interpret the target

Memory Accuracy as a Function of Prototype woman's behavior (based on her appearance,
Consistency and Timing of Occupational personality, and so on), knowledge of her
Information occupation still had an impact. Not only is
Timing of occupational occupation a potentially important category
information scheme (Feldman, 1972), but a single cat-
egory may have a noticeable processing ef-
Prototype Prior to Following fect even in an information-rich context. In
features videotape videotape
this research, subjects selectively processed
Consistent .74 .68 highly concrete information about physical
Inconsistent .66 .57 appearance and life-style rather than per-
sonality-related behaviors. Selective social
Note. Average percentage correct out of four or five processing apparently extends to objective
items.
facts about "people as well as to personality
cues.
rately remember both consistent and incon-
sistent information about her. Presumably,
these subjects used the occupational infor- Consistent and Inconsistent Information
mation as an organizing scheme that led to It is interesting that having an occupa-
the increased memorability of features that tional label in mind while encoding the vi-
were both consistent and inconsistent with gnette allowed subjects to be more accurate
their category prototype. at recognizing both consistent and inconsis-
These results indicate that both encoding tent information. Although prototype-con-
and retrieval processes contributed to selec- sistent features were better remembered
tive accuracy effects. First, the greater ac- overall, memory for prototype-inconsistent
curacy of subjects who knew the target's features was also strengthened by knowing
occupation while watching strongly suggests the woman's occupation while watching her.
that their encoding of the vignette was If the occupational label brought to mind
guided by this knowledge. Second, when oc- characteristics of perceivers' occupational
cupational information was presented after prototypes, then the processing of consistent
the tape, the selective accuracy effect could features was presumably facilitated because
only have been due to retrieval processes. they were similar to these characteristics. At
Evidence of category-based encoding and the same time, the obvious discrepancy of
retrieval in this study both replicates and inconsistent features from the prototype may
extends the results of prior research (e.g., have made these features particularly sa-
Rothbart et al., 1979; Zadny & Gerard, lient.
1974). Of course this explanation is appropriate
only if the inconsistent features were in fact
General Discussion perceived by subjects as inconsistent with
their prototypes. Although recognition items
The data presented here add to the ac- were selected to be both highly likely for one
cumulating literature about the processes prototype and highly unlikely for the other,
and boundaries of social perception, repli- it is possible that some items were simply
cating some previous results and allowing irrelevant for the nontarget prototype. If so,
some prior conclusions to be extended. Sub- then other explanations must be considered;
jects' memory for a rich behavioral sequence for example, knowledge of the target's oc-
was influenced by their abstract knowledge cupation may have led to increased attention
about the target person's occupation. As while viewing her behavior. Barring inter-
Hastie and Kumar (1979) note, category- pretational difficulties, these data appear to
based processing will be evident only when be among the first evidence that perceivers
the category manipulated by the experi- with a relevant category label were more
menter is one that is salient to the subject. accurate for category-relevant information,
Despite the various categories that subjects both consistent and inconsistent, than per-
450 CLAUDIA E. COHEN

ceivers with no such label.5 Future research may have been processed more quickly and
might systematically compare the processing thus were more likely to be encoded, given
of category-inconsistent with category-irrel- limited encoding capacity (Norman, 1976).
evant information. Second, ambiguous information might have
How does the superior category-consis- been interpreted by subjects as prototype
tent information accuracy result found in consistent. For example, although the woman
this research fit with the data from previous asked her husband to turn on the television,
studies? As suggested earlier, two factors in it was not made explicit that they spent the
the current paradigm may have promoted whole evening watching it. Waitress-set sub-
category-consistent information memorabil- jects may have interpreted this as evidence
ity: (a) a potential information overload for of heavy television viewing, whereas librar-
the perceiver and (b) a balance of consistent ian-set subjects may have concluded that she
and inconsistent features, possibly minimiz- only watched the national news.
ing the salience of the information. A third Obviously, the memory data from subjects
factor, instructions to form an impression of who were told her occupation after viewing
the target person, is associated with conflict- the tape reflect the role of retrieval rather
ing results, perhaps due to other differences than encoding processes. However, it is not
between paradigms. In Hastie and Kumar's clear whether these data should be inter-
(1979) paradigm, in which inconsistent fea- preted as evidence of selective accuracy or
tures were better remembered, subjects were whether biased reconstruction may have also
instructed both to recall the list of traits and contributed to this result. Due to the nature
behaviors describing a target person and to of the forced-choice memory measure, sub-
fcrm a personality impression of him. Recall jects may have occasionally chosen a pro-
instructions may have promoted rigorous totype-consistent feature even though they
attention to all trait and behavioral infor- did not actually remember it. This process
mation, whereas the integration of the in- might be better viewed as prototype-based
formation required by the impression in- reconstruction.6 Future research could ad-
structions may have made inconsistencies dress this processing distinction by using a
particularly salient. However, in the current different dependent measure. If two proto-
paradigm, in which consistent features were type-consistent response alternatives were
better remembered, subjects were not given included (one that did appear and one that
recall instructions. Further, the inconsisten- did not appear in the videotape), increased
cies were not in the target's personality-rel- accuracy could be more clearly distinguished
evant behavior but rather in life-style char- from biased reconstruction of the vignette
acteristics, given her occupation. This (Ostrom, Note 2).
information was probably less relevant for
forming a personality impression than were
the lists of traits and behaviors in Hastie and 5
For example, Rothbart et al. (1979) found that sub-
Kumar's (1979) study; thus, the stimulus jects who were given an expectation before a series of
inconsistencies in the current study may behaviors had superior recall for expectancy-consistent
but not expectancy-inconsistent behaviors. Also, al-
have been less salient. Knowledge of the tar- though there was a main effect of timing in their study,
get's occupation, therefore, promoted greater in which expectation-before subjects exhibited better
accuracy for prototype-consistent features. recall than expectation-after subjects, this result may
have been attributable to differences between the sub-
ject populations in the two groups.
Encoding and Retrieval Processes 'Snyder and Uranowitz (1978) found that a stereo-
type label given after subjects read a case history led
The results from Experiment 2 demon- them to choose stereotype-consistent life-style features
strated that both encoding and retrieval ef- when responding to recognition memory questions about
fects contributed to the selective memory the case history. However, their ability to accurately
result. Encoding processes may have been recognize the stereotype-consistent features actually
contained in the case history was not clearly facilitated.
influenced by prototypic knowledge in two The data seem to be evidence of biased reconstruction
ways. First, prototype-consistent features rather than selective accuracy.
 PERSON CATEGORIES AND SOCIAL PERCEPTION 451

Implications of Category-Based memory (cf. Hastie & Carlston, 1980). The

Processing development of a model of the respective
roles of prior social knowledge and novel in-
Evidence that perceivers may selectively formation should be facilitated by a better
process objective facts about people's be- understanding of perceivers' implicit social
havior has obvious relevance for many sit- knowledge.
uations. Courts of law, hiring and personnel •
decisions, classrooms, and so on all require Reference Notes
accurate memory of the personal character-
istics and behavior of others. Further, to the 1. Cohen, C. E. Prototypes: Some properties of person
extent that friendships are based on per- categories. In R. Hastie (Chair), Categorical pro-
cessing and representation of person information,
ceived life-style compatibility, individuals' symposium presented at American Psychological
personal lives may also be affected by cat- Association convention, New York, September 1979.
egory-based processing of other people's be- 2. Ostrom, T. Personal communication, November 1977.
havior.
Some tentative comments about differing References
theoretical approaches to person perception
can be made that were not possible from Ashmore, R. C., & Del Boca, F. K. Sex stereotypes and
implicit personality theory: Toward a cognitive-social
previous research. Two conflicting sets of psychological conceptualization. Sex Roles, 1979, 5,
assumptions are apparent in the current lit- 219-248.
erature. One line of reasoning argues that Bahrick, H. P., Bahrick, P. O., & Wittlinger, R. P. Fifty
realistic concrete examples of behavior should years of memory for names and faces: A cross-sec-
tional approach. Journal of Experimental Psychol-
override abstract stereotypic expectations ogy: General, 1975, 104, 54-75.
(Borgida, Locksley, & Brekke, in press; Nis- Borgida, E., Locksley, A., & Brekke, N. Social stereo-
bett & Borgida, 1975). The other argument types and social judgment. In N. Cantor & J. F.
states that when person information is com- Kihlstrom (Eds.), Cognition, social interaction, and
personality. Hillsdale, N.J.: Erlbaum, in press.
plex and lifelike, stereotypic knowledge is Brigham, J. C. Ethnic stereotypes. Psychological Bul-
useful for reducing the processing load (cf. letin, 1971, 76, 15-38.
Taylor & Crocker, 1980). The data from Cantor, N., & Mischel, W. Traits as prototypes: Effects
this study support the latter argument, since on recognition memory. Journal of Personality and
perceivers used prototypic knowledge about Social Psychology, 1977, 35, 38-48.
occupation members even when confronted Cantor, N., & Mischel, W. Prototypes in person per-
ception. In L. Berkowitz (Ed.), Advances in experi-
with vivid and concrete violations of many mental social psychology (Vol. 12). New York: Ac-
of their preconceptions. These results are ademic Press, 1979.
consistent with a view, mentioned earlier, Carlston, D. E. Events, inferences, and impression for-
that assumes person information is orga- mation. In R. Hastie et al. (Eds.), Person memory:
The cognitive basis of social perception. Hillsdale,
nized around category prototypes and that N.J.: Erlbaum, 1980.
prototypic features are used in processing Cohen, C. E. Goals and schemata in person perception:
new information about individual category Making sense from the stream of behavior. In N.
members. Cantor & J. F. Kihlstrom (Eds.), Cognition, social
interaction, and personality. Hillsdale, N.J.: Erl-
Given that observers do rely on their im- baum, in press.
plicit knowledge about categories of people, Cohen, C. E., & Ebbesen, E. B. Observational goals and
the range of these categories and the infor- schema activation: A theoretical framework for be-
mation they contain should be investigated havior perception. Journal of Experimental Social
further. Some recent work by Cantor and Feldman, Psychology, 1979, 15, 305-329.
J. M. Stimulus characteristics and subject
Mischel (1979) demonstrates that people prejudice as determinants of stereotype attribution.
can describe both the personality and phys- Journal of Personality and Social Psychology, 1972,
ical characteristics associated with many 21, 333-340.
category prototypes. Further assessment of Goldstein, A. G. & Chance, J. E. Visual recognition
memory for complex configurations. Perception &
category prototypes and taxonomies of cat- Psychophysics, 1971, 9, 237-241.
egories is needed for a better understanding Hamilton, D. L. A cognitive-attributional analysis of
of the organization of person information in stereotyping. In L. Berkowitz (Ed.), Advances in Ex-
452 CLAUDIA E. COHEN

perimental Social Psychology (Vol. 12). New York: nance of social stereotypes. Journal of Experimental
Academic Press, 1979. Social Psychology, 1979, 15, 343-355.
Hastie, R. Memory for behavioral information that con- Rothbart, M., Fulero, S., Jensen, C., Howard, J., &
firms or contradicts a personality impression. In R. Birrell, P. From individual to group impressions:
Hastie et al., (Eds.), Person memory: The cognitive Availability heuristics in stereotype formation. Jour-
basis of social perception. Hillsdale, N.J.: Erlbaum, nal of Experimental Social Psychology, 1978, 14,
1980. 237-255.
Hastie, R., & Carlston, D. Theoretical issues in person Shepard, R. N. Recognition memory for words, sen-
memory. In R. Hastie et al. (Eds.), Person memory: tences and pictures. Journal of Verbal Learning and
The cognitive basis of social perception. Hillsdale, Verbal Behavior, 1967, 6, 156-163.
N.J.: Erlbaum, 1980. Snyder, M., & Uranowitz, S. Reconstructing the past:
Hastie, R., & Kumar, P. A. Person memory: Personality Some cognitive consequences of person perception.
traits as organizing principles in memory for behav- Journal of Personality and Social Psychology, 1978,
iors. Journal of Personality and Social Psychology, 36, 941-950.
1979, 37, 25-38. Taylor, S. E., & Crocker, J. Schematic bases of social
Jeffery, K. M., & Mischel, W. Effects of purpose on information processing. In E. T. Higgins, C. P. Her-
the organization and recall of information in person man, & M. P. Zanna (Eds.), The Ontario Symposium
perception. Journal of Personality, 1979, 47, 397- on Personality and Social Psychology: Social Cog-
419. nition. Hillsdale, N.J.: Erlbaum, 1980.
Nisbett, R., & Borgida, E. Attribution and the psy- Taylor, S. E., Fiske, S. T., Etcoff, N. L., & Ruderman,
chology of prediction. Journal of Personality and A. J. Categorical and contextual bases of person mem-
Social Psychology, 1975, 32, 932-943. ory and stereotyping. Journal of Personality and So-
Norman, D. A. Memory and attention (2nd ed.). New cial Psychology, 1978, 36, 778-793.
York: Wiley, 1976. Tulving, E., & Thomson, D. M. Encoding specificity
Palmer, S. E. Visual perception and world knowledge: and retrieval processes in episodic memory. Psycho-
Notes on a model of sensory-cognitive interaction. In logical Review, 1973, 80, 352-373.
Norman et al., Explorations in cognition. San Fran- Zadny, J. & Gerard, H. Attributed intentions and in-
cisco: Freeman, 1975. formational selectivity. Journal of Experimental So-
Rosch, E. Cognitive representations of semantic cate- cial Psychology, 1974, 10, 34-52.
gories. Journal of Experimental Psychology: Gen-
eral, 1975, 104, 192-233.
Rothbart, M., Evans, M., & Fulero, S. Recall for con- Received January 18, 1980
firming events: Memory processes and the mainte- Revision received September 2, 1980