Journal of Personality and Social Psychology

1977, Vol. 35, No. 9, 677-688

Self-Reference and the Encoding of Personal Information

T. B. Rogers, N. A. Kuiper, and W. S. Kirker
University of Calgary, Canada

The degree to which the self is implicated in processing personal information

was investigated. Subjects rated adjectives on four tasks designed to force
varying kinds of encoding: structural, phonemic, semantic, and self-reference.
In two experiments, incidental recall of the rated words indicated that adjec-
tives rated under the self-reference task were recalled the best. These results
indicate that self-reference is a rich and powerful encoding process. As an
aspect of the human information-processing system, the self appears to func-
tion as a superordinate schema that is deeply involved in the processing, inter-
pretation, and memory of personal information.

Present research and theory in personality and how these explanations are organized in
appear to be placing more and more empha- an attributional network. The common
sis on how a person has organized his or her thread in all of these contemporary research
psychological world. Starting with Kelly's areas is the notion that the cognitions of a
(1955) formulation of personal constructs, person, particularly their manner of organ-
we see a gradual emergence of a number of ization, should be an integral part of our
avenues of inquiry that use this as their focal attempts to explain personality and behavior.
point. In person perception, the concept of Of concern in the present article is the
lay personality theory stresses that the ob- construct of self and how it is implicated in
server's analytic network of expected trait the organization of personal data. Our gen-
covariations is an integral part of how he eral position is that the self is an extremely
processes (and generates) interpersonal data active and powerful agent in the organization
(Hastorf, Schneider, & Polefka, 1970). Bern of the person's world. More specifically, the
and Allen (1974), in their embellishment of present research was designed to determine
Allport's (1937) idiographic position, argue if self-reference serves a meaningful function
that an important determinant of predictive in the processing of certain kinds of informa-
utility of trait measurement is the manner tion. That is, we attempted to determine the
in which the respondent has organized his or relative strength of self-reference as an agent
her view of the trait being measured. These in the processing of people-related informa-
authors see the overlap between the respond- tion.
ent's and the experimenter's concept of the The self is defined as an abstract repre-
trait as a necessary prerequisite of predic- sentation of past experience with personal
tion. Attribution theory (Jones et al., 1971) data. Phenomenologically, it is a kind of vague
is another example of this increased accent idea about who the person thinks he or she
on personal organization. Here the emphasis is. It probably develops to help the person
is on how the subject explains past behavior keep track of the vast amounts of self-rele-
vant information encountered over a lifetime.
The self, then, represents the abstracted es-
This research was supported by a grant from the
Canada Council. We would like to thank the fol- sence of a person's perception of him or her-
lowing persons for their useful ideas and comments self. A more formal definition of self is to
on earlier drafts: F. I. M. Craik, E. J. Rowe, P. J. view it as a list of terms or features that have
Rogers, H. Lytton, J. Clark, J. Ells, C. G. Costello, been derived from a lifetime of experience
and especially one anonymous reviewer.
with personal data. More than likely a por-
Requests for reprints should be sent to T. B.
Rogers, Department of Psychology, The University tion of the list consists of general terms—
of Calgary, Calgary, Alberta, Canada, T2N 1N4. not unlike traits—that represent the ab-
677
678 T. ROGERS, N. KUIPER, AND W. KIRKER

stracted essentials of a person's view of self. between the previous experience of the in-
In addition to these general terms, there are dividual (in the form of the abstract struc-
also some more specific entries in the self. ture of self) and the incoming materials. The
These relate to less salient and more situa- involvement of the self in the interpretation
tion-specific aspects of self-perception as of new stimuli imparts a degree of richness
well as to specific behaviors. This definition and fullness to the input because of the avail-
is supported by Jones, Sensening, and Haley ability of the immense amounts of previous
(1974). They simply asked subjects to de- experience embodied in the self.
scribe their "most significant characteristics." The interaction between new input and
The most frequent entries in the obtained previous experience, postulated to be central
response protocols were positively worded to self-reference, has been modeled in the
terms such as sensitive, intelligent, and cognitive literature under the concept of
friendly. These appear to be the general schema or prototype (Bartlett, 1932; Posner
terms in the self and appear to resemble & Keele, 1968). For example, subjects shown
traits. Jones et al. (1974, p. 38) also present a series of dot patterns tend to abstract a
a single response protocol. Of interest here prototypical visual pattern and use it as a
is the tendency for conditionals (e.g., such standard in a memory task (Posner & Keele,
situational hedges as sometimes) to emerge 1970). In personality, several recent papers
later in the protocol. Furthermore, as the have suggested that personal data are pro-
protocol develops, the entries tend to relate cessed using schemata or prototypes. Markus
to more specific situations than did the earlier (1977), using a series of measurements, cate-
terms (e.g., "have a hang-up about authority gorized subjects by whether dependence was
figures"). These latter entries appear to be part of their self. Schematics, or those with
the specific terms in the self. dependence as part of their self, were those
One of the main functions of the self is who rated themselves as extreme on several
to help the individual process personal data. dependence items, as well as indicating that
When a person encounters a situation involv- they viewed dependence as important. As-
ing personal information, this structure is chematics were midrange on dependence and
activated and becomes part of the available low on importance ratings, representing sub-
information-processing system. For example, jects who did not have dependence in their
when students encounter a list of character- general concept of self. On the basis of im-
istics of a psychopathological state (e.g., in pressive convergent evidence, Markus (1977)
an introductory psychology lecture), they found that schematics and aschematics
tend to interpret (and attempt to remember) showed differences in how they processed per-
these by referring them to their own views of sonal data. For example, schematics were
self. Such a strategy could lead to the "medi- more resistant to incorrect personal informa-
cal student syndrome," where students begin tion than were the aschematics. These data
to see themselves in the varying states de- suggest that the traits, such as those found
scribed by the lecturer. In extreme cases, in the self, serve an important function in
some students can be convinced they are processing certain kinds of information.
raving lunatics—despite repeated warnings Cantor and Mischel (1977) tested the
of the instructor. Other examples of this self- proposition that traits function as prototypes
reference phenomenon can occur in situations using a recognition memory task. Subjects
involving personal feedback, expressive be- were shown a series of statements that rep-
haviors, and the processing of information resented an introvert. When faced with a
about other people. task requiring recognition of these statements
The central aspect of self-reference is that from among some new introvert statements,
the self acts as a background or setting subjects tended to misidentify some of the
against which incoming data are interpreted new items as having been original state-
or coded. This process involves an interaction ments. This suggests that the concept of in-
 SELF-REFERENCE AND ENCODING 679

trovert mediated subjects' processing of the To review, self-reference can be seen as a

initial set of statements. This mediation was process involving the schema of self. This
postulated to be in the form of a prototype, process involves the interaction between pre-
which represents an abstraction of the con- vious experience with personal data and new
cept of introvert from the initial items. The stimulus input. When self-reference is in-
observed memory bias for new, yet conceptu- volved, it should provide a useful device for
ally related items reflected the involvement encoding or interpreting incoming informa-
of this abstraction. tion by virtue of accessing the extensive past
The Markus (1977) and Cantor and Mis- experience abstracted in the self. Contact
chel (1977) data indicate that traits are im- with the reservoir of history embodied in the
plicated in information-processing functions. self should provide considerable embellish-
They appear to be involved in the organiza- ment and richness to an incoming stimulus.
tion, storage, and retrieval of personality- Rogers (in press) explored this possibility
related information. Our view of self places using recognition memory for personality
these traitlike schemata or prototypes as gen- items. Subjects instructed to "read the item,
eral terms in the feature list making up the decide if it describes you, and use this to
self. help your memory" performed significantly
Rogers, Rogers, and Kuiper (Note 1) ex- better than subjects receiving either no or
plored the manner in which this set of different (i.e., imagery) instructions. These
schemata is involved in processing personal data, in combination with some older re-
data. They hypothesized that the self func- search (e.g., Cartwright, 1956), indicate that
tions like a grand or superordinate schema. explicit instructions to use the self in a
If the self is a schema, it should be possible memory task increase performance, which
to observe the kinds of memory biases docu- supports the claim that self-reference serves
mented by Cantor and Mischel (1977). In to enrich input.
one study (Rogers et al., Note 1, Experiment This enriching aspect of self-reference is
2), subjects filled out self-ratings on 84 ad- the focal point of the present article. Our
jectives. Two and a half months later, these major concern is how powerful self-reference
same subjects participated in a recognition is as an encoding device. More specifically,
memory study involving these same adjec- self-reference is compared to several other
tives. They first saw a randomly selected set encoding processes in an effort to determine
of 42 of the words, and then had to recognize the relative degree of richness and embellish-
these from among the total set of 84. If the ment that self-reference imparts to the en-
self was involved as a schema, subjects coding of adjectives.
should tend to falsely recognize new items The experimental manipulation used in the
that were rated as self-descriptive (i.e., Pos- present context is an incidental recall para-
ner & Keele, 1970). Each subject's recogni- digm, in which subjects make different kinds
tion protocol was divided into high, neutral, of ratings on a set of words. For example, a
and low self-descriptive categories on the subject rates whether a given word means the
basis of their self-ratings. Performance for same as a target word. This would be a
the 42 items initially shown in the recogni- semantic rating, as the subject must extract
tion study was not affected by degree of the meaning of the word to perform the task.
self-reference. However, performance on the The same subject rates whether another word
new or distractor items (correct rejects) be- is written in big letters. This is a structural
came poorer as degree of self-reference in- coding task, since all the subject has to do
creased. In other words, more false alarms is inspect the structure of the stimulus item
occurred as the adjectives became more self- rather than extract the meaning of the word.
descriptive. This clearly confirms the predic- Other words are rated on a phonemic task,
tion derived from viewing the self as a which involves deciding whether a word
schema. rhymes with a target. The self-reference rat-
680 T. ROGERS, N. KUIPER, AND W. KIRKER

ing, which subjects perform on some of the parison permits assessment of the degree to
words, involves the respondent's deciding which the self aids in producing a stronger
whether the word describes him or her. When trace, in contrast to usual semantic encoding.
the rating task is completed, each subject has If the self is an active agent in the encoding
rated one fourth of the words on each of the of personal data, we predicted that the self-
four rating tasks. These four tasks are reference rating would produce good inci-
thought to vary in depth, or semantic rich- dental recall in this depth-of-processing para-
ness, from the structural task as the most digm. If incidental recall of the self-refer-
shallow to the semantic and/or self-rating ence words is superior to that for semantic
task as the deepest. The test of coding words, the hypothesis that the self serves
strength comes when the subjects are given an active and powerful role in processing
a surprise recall task at the conclusion of personal data would be supported.
the ratings. According to Craik and Lock- The present article offers two experiments
hart (1972), words that have been deeply that examine this proposition. The first study
coded during the rating task should be re- involves a close replication of Craik and
called better than words with shallow coding. Tulving's (1975) initial experiments, with
This manipulation permits us to determine the self-reference task included. The second
the relative deepness of self-reference as a experiment replicates and extends the first
coding device. study by using a different technique and
Craik and Tulving (1975) have done a different semantic rating task.
series of studies using this methodology.
They have restricted their efforts to the Experiment 1
structural, phonemic, and semantic types of
tasks. Their results indicate that recall (or This experiment is intended to determine
recognition) is best for semantic tasks and the relative position of self-reference in
poorest for structural ratings, with phonemic Craik's (Craik & Lockart, 1972) depth hier-
in the middle. These data are interpreted as archy. The procedural details have been
support for the position that the strength of chosen to closely approximate Craik and Tul-
the memory trace is "a positive function of ving's (1975) initial experiments in an effort
'depth' of processing, where depth refers to to maximize the degree of comparability of
greater degrees of semantic involvement" the present results.
(Craik & Tulving, 1975, p. 268). Presuma-
bly the rating tasks (structural, phonemic, Method
etc.) force the subject to code the word to a
The study has two main parts. First, subjects
specific level, and the incidental recall is a rated 40 adjectives on one of four tasks. This in-
function of the depth of these tasks. These volved presenting a cue question, followed by 1
kinds of results have emerged quite consist- of the 40 adjectives. Subjects answered yes or no
ently in the cognitive literature (e.g., D'Agos- to the cue question as it applied to the adjective.
tino, O'Neill, & Paivio, 1977; Klein & Saltz, The cue questions, along with the manipulations
for each task, are presented in Table 1. After com-
1976; Schulman, 1974; Walsh & Jenkins, pleting the ratings, subjects attempted to recall the
1973). adjectives in the second part of the study.
Of particular concern in the present study Materials. The main items for this study were
40 adjectives that were deemed appropriate for a
is the comparison between incidental recall self-description task. They were chosen to represent
for words rated under the semantic and self- a broad spectrum of possible characteristics and
reference tasks. Both of these tasks involve were selected from all of the trait descriptions found
in Jackson's (1967) Personality Research Form A
semantic encoding, but there is an important Manual. Thirty-eight of the adjectives, selected to
difference between them. The self-reference be familiar to the subject population, came from
task forces the subject to use the self in this source. Two other adjectives (shy and out-
going) were added to make up the total of 40.
the rating task, whereas the semantic task We used Roget's Thesaurus to construct a further
does not. The self-reference/semantic com- set of 40 synonyms for the semantic tasks. The
 SELF-REFERENCE AND ENCODING 681

Table 1
Examples of the Rating Tasks

Task Cue question Manipulation

Structural Big letters? The adjective was either presented in the

same size type as the question or twice
as large.
Phonemic Rhymes with x x x x ? x x x x was a word that either rhymed or
did not rhyme with the adjective.
Semantic Means same as YYYY? YYYY was either a synonym or unrelated
word to the presented adjective.
Self-reference Describes you ? Subjects simply responded yes or no to
indicate the self-reference quality of the
presented adjective.

final synonyms chosen represented consensus among tial instructions did not indicate that recall was
the three authors. expected. All stimuli were displayed on a television
The phonemic task dictated a second supple- monitor driven by a PDP8/1 computer, which also
mentary list of 40 words that rhymed with the recorded the ratings and rating times. Including the
main adjective set. The authors generated a set of four buffer items at the end and beginning of the
possible rhyming words, and consensus among our- list, there were 48 rating trials. Each of these con-
selves was the final criterion for selection. Most sisted of (a) a 3-sec presentation of the cue ques-
(90%) of these words were adjectives. tion, (b) a 500-msec blank interval, (c) presentation
A third supplementary list of nonsynonym, non- of the target adjective, which was terminated by
rhyming words was also required, so that one half the subject's response, indicated on a two-button
of the cue questions could result in a no rating. response panel placed comfortably in front of the
Kirby and Gardner's (1972) set of adjectives was subject, and (d) a 2-sec intertrial interval before
consulted to derive this list. Again, author con- the next cue question was presented. After the rat-
sensus regarding the nonrhyming and nonsynonym ing task, the subject was given a piece of paper
quality of the adjectives dictated the final list. and was asked to recall, in any order, the adjectives
A set of eight further adjectives and supplemen- he or she had rated. Three minutes were allowed for
tary words was generated to provide buffer items recall.
of four ratings each at the beginning and end of Subjects. Volunteers from the introductory psy-
the list. These items, which were constant across chology subject pool served as subjects. There were
lists, were not included in the data analysis. This 32 subjects (16 female and 16 male) with an average
was intended to minimize the effects of primacy and age of 20.2 years. Each was paid $1.50 for par-
recency in the incidental recall task. ticipating. Subjects were randomly assigned to the
Four lists of adjectives were constructed, such eight list conditions, yielding 4 subjects per order. .
that 10 adjectives in each list were rated under
each cue question, and over the four lists, each
adjective was rated under each cue question.
Results and Discussion
To guard against the possibility that wo-rated
words are recalled differently than yes-rated words,
For each subject, the number of adjectives
each of the four lists was reversed to generate eight recalled as a function of rating task (struc-
lists in total. For example, if in a given list, under tural, phonemic, etc.) and observed yes or no
the structural task, a word appeared in small letters rating was calculated. The means of these
(generating a no response), the reversed list would figures are presented in the top panel of
have the word presented in big letters (generating
a yes response). The one exception to this counter- Table 2. A 4 (rating tasks) X 2 (yes/no rat-
balancing was the self-reference task. Here it was ing) two-way analysis of variance revealed
impossible to have experimental control over yes a significant main effect of rating task, F(3,
and no responses, since the person's view of self 93) = 29.01, p < .001. Newman-Keuls tests
would dictate his or her response.
indicated meaningful differences (p < .05 or
In all lists, order of the cue questions was ran-
domly assigned in blocks of eight trials, such that
better) in the recall for all points in this
each combination of cue question and expected main effect except for the structural-pho-
response was represented once every eight trials. nemic comparison. The main effect of rating
Procedure. Subjects were tested individually. Ini- was also significant, F(l, 31) = 4.22, p<
682 T. ROGERS, N. KUIPER, AND W. KIRKER

.05, indicating superior recall for words given mata that are part of the self (see Rogers,
a yes rating. The Rating Task X Rating in- 1974). This comparison culminates in the
teraction was also significant, F(3, 93) = subject's yes/no response, which leads to a
3.47, p < .05. Post hoc tests of this inter- strong and specific encoding of the rated
action revealed a meaningful yes/no differ- item. During the recall phase of the study,
ence for the self-reference rating, £(31) = items with this detailed and specific encoding
2.62, p < .05. are easily retrieved, producing good recall
The overall pattern of these results is simi- performance.
lar to that typically found in the literature In the semantic rating task it is unneces-
(e.g., Experiments 1, 2, and 3 in Craik & sary for an elaborate structure such as the
Tulving, 1975). The main effect of rating self to be involved. Rather, the subject ac-
task is used to suggest that the depth to cesses his associative memory (e.g., Estes,
which items are processed during the rating 1976) for the target adjective and makes his
task determines the strength of the memory synonymity judgment from this. The result-
trace. As recall is a function of trace strength, ing trace is not as specific or detailed as that
the present results support this position. involved with self-reference. Clearly, the ac-
Of central interest is the finding that the cess of associative memory produces a more
self-reference task develops a stronger trace detailed trace than either the structural or
than the semantic task, as shown by the sig- phonemic tasks. However, when compared to
nificant recall differences between these two self-reference, the trace derived from a syno-
conditions. This result clearly supports the nymity judgment is relatively weak. This
idea that self-reference functions as a pow- difference in the specificity of the self-refer-
erful coding device. In the case of self-refer- ence and semantic tasks seems to be the
ence ratings, the subject uses his or her con- major reason for the inferior incidental re-
cept of self to respond to the adjective. The call of the semantically rated words.
self-ratings involve comparison of the in- The time required to make the ratings is
coming adjective with the terms and sche- typically used to monitor the effectiveness

Table 2
Recall, Rating Time, and Adjusted Recall as a Function of Rating Task and Rating for
Experiment 1

Rating task

Rating Structural Phonemic Semantic Self-reference

Mean recall Total

yes .28 .34 .65 1.78 3.05
no .06 .34 .68 1.06 2.14
Total .34 .68 1.33 2.84 5.19

Mean rating time (msec) M

yes 1,267 2,177 2,255 3,194 2,223
no 1,474 2,104 3,006 2,689 2,318
M 1,371 2,141 2,631 2,941 2,271

Mean adjusted recall M

yes .05a .08 .14 .30 .14
no .01 .06 .12 .29 .13
M .03 .07 .13 .30 .13
a
This figure represents the mean (over 32 subjects) of the number of recalled ^5-rated structural items
divided by the number of yes ratings made on the structural task.
 SELF-REFERENCE AND ENCODING 683

of the experimental manipulations in this each subject, the number of yes responses
paradigm. Further, these reaction time (RT) made under each rating task (maximum =
data provide convergent evidence for self- 10) was calculated. The means were 5.00,
reference as a useful encoding task. The RTs 4.34, 4.06, and 6.13 for the structural, pho-
from the present study were sorted separately nemic, semantic and self-reference tasks, re-
for each subject into yes/no by rating-task spectively. A simple analysis of variance on
categories. The means are presented in the these figures revealed a significant effect,
middle panel of Table 2. Only the main effect F(3, 93) = 16.99, p < .001, indicating that
of rating task was significant in this analysis, number of yes responses is related to rating
F(3, 93) = 10.35, p < .001. A clear linear task. The deviations from 50% yes responses
trend in RTs is evident in these data, indi- for the phonemic and semantic tasks are due
cating maximal RT for the self-reference to the difficulty of constructing exact rhymes
rating task.1 This analysis replicates Craik and synonyms for the adjectives.
and Tulving (1975) and is clearly compat- More important than the significant varia-
ible with the recall data presented above, tion in number of yes responses is the pos-
supporting the involvement of the self as a sible effect this might have on the recall data.
coding device. Since it is already known that yes-rated
The finding that yes-rated words are re- words are better recalled (e.g., Craik &
called better than «0-rated words occurs in Tulving, 1975), it is possible that self-refer-
other studies and has a number of interesting ence recall was superior because subjects
implications. Craik explains these data by made more yes responses in the self-reference
arguing that in the case of yes-ra,ted words, task. To assess this, the recall data were
the "encoding questions or context forms an transformed to a proportion score that ad-
integrated unit with the target word" (Craik justs for differential numbers of yes re-
& Tulving, 1975, p. 291). Presumably this sponses. Specifically, a particular subject's re-
integrated unit forms a stronger trace than call of ;yej-rated words under a specific rating
less integrated ones (no-rated words), task was divided by the number of yes rat-
thereby augmenting recall. The interaction ings the subject made while doing the task.
observed in the present data indicates that This transformed score represents the pro-
this yes/no difference occurred only for the portion of recalled words the subject rated
self-reference case, which suggests that items as yes. Similarly, the wo-rated word recall
viewed as self-descriptive (yes-rated words) under a given rating task can be divided by
form a "more integrated unit" than do the number of no responses made on this rat-
non-self-descriptive terms. These results
strengthen even more our view of self, as it
appears that terms that match the subject's 1
It is possible that items with large RTs are
self-view become more integrated than those better recalled, calling into question this interpreta-
that do not match. This finding is consonant tion. If study time is the important factor, its
with both Markus's (1977) and Cantor and effects should be observable within each task as
well as across tasks. Thus, within a given rating
MischePs (1977) finding that personal data task, the items with the longer study times should
are processed using schemalike structures. be recalled better. To explore this, the 10 RTs
There are several aspects of these data under each of the four rating tasks were subdivided
separately for each subject into a fast and a slow
that require examination before the previous subset (5 RTs in each). The recall for these sub-
conclusions are fully warranted. The yes/no sets was analyzed in a 4 (rating tasks) X 2 (fast
difference in recall for self-reference words and slow study times) two-way analysis of variance.
The study-time hypothesis predicts significant effects
could be due to a differential number of yes for the terms involving study time. The analysis in-
responses as a function of rating task. Since dicated only the expected main effect of rating task,
experimental control over the number of yes F(3, 93) =30.85, p < .001. This analysis weakens
the study-time interpretation and reinforces the
responses was not possible for the self-refer- interpretation that the recall data are due to the
ence task, this is a distinct possibility. For qualitative nature of the various encoding tasks.
684 T. ROGERS, N. KUIPER, AND W. KIRKER

ing task to provide a score representing the tasks can be thought of as a family of judg-
proportion of recalled words rated as no. ments, all of which involve the extraction
Note that this is a subject-specific correction (and possibly some elaboration) of the mean-
that reflects recall corrected for differential ing of the target item. Such tasks as syno-
numbers of yes and no ratings. The means of nymity ratings, judgments of semantic speci-
the adjusted recall scores are presented in ficity, and deciding whether a word fits into
the bottom panel of Table 2. An analysis of a sentence frame can be considered members
variance of these data revealed only a main of this semantic family. Experiment 2 em-
effect of rating task, F(3, 93) =31.63, p < ployed meaningfulness ratings as the semantic
.001. The important recall difference between task. Since recall is a function of meaning-
semantic and self-reference survived this ad- fulness (see Noble, 1952; Paivio, Yuille, &
justment (p<.0l), but the yes/no differ- Rogers, 1969), this encoding task should be
ence for the self-reference task did not. This very beneficial for recall, particularly for
analysis reaffirms self-reference as a coding words given a yes rating. If self-reference
tool but questions the possibility that yes- emerges as superior to meaningfulness rat-
rated items form a more integrated unit. ings, evidence confirming the strength and re-
In summary, the data from Experiment 1 liability of self-reference as an encoding de-
provide evidence that self-reference is a pow- vice will be provided.
erful encoding device. The superior incidental A second purpose of this experiment is to
recall of adjectives rated under the self-refer- explore the robustness of the self-reference
ence task, in combination with the RT data, findings. Experiment 1 was performed using
suggests that self-reference provides a rich fairly tight experimental controls. The pres-
and powerful encoding. The involvement of ent experiment deviates from this by using
self in the rating task provides a good en- a group testing procedure. Craik and Tulving
coding unit, which functions effectively as a (1975) and Klein and Saltz (1976) have
memory cue. used similar procedures and replicated the
findings from more rigorous paradigms, sug-
Experiment 2 gesting that the self-reference finding should
stand up in this group procedure.
It is possible that the superiority of self-
reference encoding documented in Experi-
ment 1 is specific to synonymity ratings. Method
Maybe other kinds of semantic tasks would The four rating tasks used for this experiment
produce equally powerful results. Semantic are outlined in Table 3. Subjects were given a

Table 3
Rating Tasks and Mean Adjusted Recall for Experiment 2

Mean adjusted recall

Yes No
Rating task Cue question Definition rating rating M

Structural Long? Rate whether you feel the word .21 .18 .20
is long or short.
Phonemic Rhythmic? Rate whether you feel the word .20 .18 .20
has a rhythmic or lyrical sound.
Semantic Meaningful? Rate whether you feel the word .23 .15 .19
is meaningful to you.
Self-reference Describes you ? Indicate whether the word .33 .31 .32
describes you.
M .24 .21 .23
 SELF-REFERENCE AND ENCODING 685

rating- sheet which indicated which of the four to produce recall that is superior to any other
tasks they were to perform on a given word; this task ever used in the incidental recall para-
was indicated by the cue questions from Table 3.
After the subjects had read the task cue to them- digm. This by itself attests to the power of
selves, an adjective was read aloud by the experi- self-reference.
menter, and the subjects made their yes or no re-
sponses on the sheet. After the ratings, subjects
turned over their rating sheets and attempted to General Discussion
recall the adjectives.
Materials. The 40 adjectives used in Experiment As a test of encoding strength, the depth-
1 made up the target items in this study. Four of-processing paradigm forces the subject to
different task orders were generated, such that each
adjective was rated under each task considered process stimuli to a specific depth by having
across the four orders, and within each order one the subject rate the words on different tasks.
fourth of the words were rated under each task. During the rating task a memory trace of the
Within each list the order of tasks was randomized rated word is created. Tasks that are deep or
in blocks of four, such that each task was repre-
sented once in every four trials.
semantically rich produce strong traces,
Procedure. Subjects were run in one group. After which in turn serve as useful cues in the
instructions, the experimenter read the item number, incidental recall of the rated words. The rela-
said the word task (which cued the subjects to read tive power of an encoding device is corre-
the cue question), and then read out the adjective. lated with incidental recall in this paradigm.
After 40 such trials, subjects were given 3 minutes
to recall, in any order, as many of the adjectives as The present data indicate unequivocally that
they could. Subjects were not expecting this free- words rated under the self-reference task
recall task. show superior recall. This indicates that self-
Subjects. Twenty-seven students in a fourth- reference represents a powerful and rich en-
year summer class served as subjects. The mean age
was 27.7 years.
coding device. Clearly, self-reference pro-
duces a rich encoding unit that can function
effectively during information processing (see
Results and Discussion also Markus, 1977; Rogers, in press; Rogers
et al., Note 1).
For each subject, recall as a function of
The major difference between the semantic
rating task and rating was calculated. These
and self-reference encoding tasks lies in the
figures were converted to adjusted recall
involvement of self in the latter rating. The
scores following the procedures for Experi-
self is a superordinate schema that contains
ment 1. The means of the adjusted recall
an abstracted record of a person's past ex-
scores are presented in the far right columns
perience with personal data. The richness of
of Table 3. Analysis of variance of the ad-
self-reference encoding shown in the present
justed recall scores produced a lone signifi-
article is due to the access of this schema.
cant main effect of rating task, F(3, 78) =
The semantic rating task does not force in-
4.20, p < .01, and a meaningful semantic/
volvement of a powerful schema, and hence
self-reference recall difference (p < .05).
fails to induce as powerful and rich an en-
Experiment 2 demonstrates self-reference
recall superiority when a meaningful rating
task is used. This kind of task has been pre- 2
The same pattern of results has been replicated
viously implicated in recall, resulting in a twice for this group procedure using different
seemingly powerful semantic encoding task. semantic rating tasks. Typically this group proce-
dure fails to replicate Craik and Tulving's (1975)
However, the present results indicate that findings for the structural, phonemic, and semantic
self-reference still is the more useful encod- tasks (see Table 2). This is probably due to the
ing task in this paradigm.2 rating tasks used. For example, in the structural
Taken in total, the results of these ex- task some subjects may have rated whether the
word was a "big" word (rather than long), which
periments indicate that self-reference induces
would be a semantic task. Regardless of this prob-
superior incidental recall compared to a di- lem, the important semantic/self-reference difference
versity of strictly semantic rating tasks. The clearly emerged in all studies using the group pro-
important thing is that self-reference appears cedure.
686 T. ROGERS, N. KUIPER, AND W. KIRKER

coding unit. The mere act of making a self- sonable to think that this hierarchy relates
referent decision produces such powerful in- to extremity. That is, the initial trait in the
ternal reactions that the attending memory self would be a person's most extreme trait,
trace is stronger than any observed in the followed by the second most extreme, and so
present experimental situation. on, until meaningful traits for the person
In order for self-reference to be such a were exhausted. This simple extremity or-
useful encoding process, the self must be a ganization may hold for some persons, but
uniform, well-structured concept. During the another aspect of the traits must also be con-
recall phase of the study, subjects probably sidered. Specifically, the salience of a trait
use the self as a retrieval cue (e.g., Mosco- for a person also adds to the organization of
vitch & Craik, 1976). In order for this to self. For example, a person who views him
be functional, the self must be a consistent or herself in the midrange on "friendly,"
and uniform schema. This property of the may perceive this characteristic as excep-
self is also shown by Rogers et al. (Note 1), tionally important, and thereby have it in-
who were able to predict memory perform- cluded in the self. Markus (1977) included
ance with a measure of self taken 2\ months this consideration when she used importance
earlier. The present data support the con- ratings in her definition of schematics. This
tention that the self is a well-structured and aspect of traits or constructs has also been
powerful schema. Presumably the self-refer- discussed under the labels of centrality
ent decision activates the superordinate (Snygg & Combs, 1949) and salience (Jack-
schema of self as well as the salient sub- son, Note 2). The important thing to note is
schemata. The strength of the trace devel- that the traits involved in a person's self are
oped from this activation suggests that a not necessarily the most extreme ones.
consistent and well-structured schema under- Rather, they represent a mix of salience and
lies these decisions. This consistency pro- extremity.
duces a rich and effective encoding unit, The inclusion of specific behaviors in the
which accounts for the present data. self derives from the work in cognition. Pos-
The present data permit some further ner and Keele (1968) postulate that a per-
statements about the schema properties of son stores both the prototype and some in-
the self. Considering the four experiments dication of how a given stimulus deviates
reported here (including the two supplemen- from this norm. This means that a schema,
tary studies in Footnote 2), it has been con- by virtue of its abstract property, must also
sistently found that yes-rated items are contain specific data indicating aspects of
better recalled than »0-rated items in the the previous input that do not conform to
self-reference task.3 These data support the the abstracted structure.
schema view, since ;yes-rated items would fit In sum, the self contains a set of ordered
more easily into the schema, and thereby features. The ordering appears to be from
form a more integrated encoding unit (Schul- general to specific, with the general terms
man, 1974). Such a pattern of results is com- (e.g., traits) ordered by a combination of
patible with Markus (1977) and Cantor and salience and extremity. The general terms
Mischel (1977) and further reinforces a view can serve as schemata when studied inde-
of the self as a schema. pendently of a person's idiographic view of
The data indicating that the self is a self (e.g. Cantor & Mischel, 1977; Markus,
schema prompt consideration of how the vari- 1977).
ous traits (i.e., subschemata) and specific The present data stress how the self can
elements (i.e., individual behaviors) are or-
ganized within this structure. The terms of ^Statistical significance for this difference tends
the self are organized in an hierarchical to disappear in the adjusted recall analyses. How-
ever, the consistent emergence of the effect across
fashion, with the most central traits repre- this series of experiments (even in the adjusted re-
sented initially. At first blush, it seems rea- call data) suggests a substantial effect.
 SELF-REFERENCE AND ENCODING 687
become involved in the encoding of personal sistency issue. As noted by Bern and Allen
data. Focusing on the organizational proper- (1974), our intuitions tell us that there are
ties of the self is by no means new (e.g., cross-situational consistencies in behavior,
Bertocci, 1945; Gergen, 1971). In fact, as even though the research data do not tend to
early as James (1890) the self has been confirm this. Our process approach would
postulated to be an active agent in the over- interpret this as follows: (a) Personal data
all human cognitive apparatus. The present are processed using the self (e.g., Experiment
research adds to this tradition by providing 1). (b) The self induces people to view novel
a strong empirical test of a proposition deriv- but self-relevant data as having been previ-
ing from such a view of self. Our emphasis is ously experienced (Rogers et al., Note 1).
upon the information-processing functions of (c) Therefore new personal data will appear
the self, specifically relating to self-reference to conform to expectation (i.e., fit into the
as an encoding device. This represents a neo- scheme), which could produce a perception
mentalistic approach (Paivio, 1975) to the of consistency. This interpretation focuses
self. While behavioral evidence (i.e., memory upon the organizational and biasing aspects
performance) is the key element in this ap- of the human information processor, which
proach, the focal concern is upon the inferred provides an alternate construction of these
construct—in this case, the self. important data. Although the utility of this
It should be noted that there are certain interpretation rests on further empirical tests,
classes of information likely to receive self- the amenability of such a model to direct ex-
reference encoding. Only after certain con- perimental scrutiny argues in its favor.
textual information indicates that the self In summary, the present article offers data
may be a functional aid to processing will to suggest that self-reference is a very potent
the schema be activated. In the present con- encoding device. The pattern of results indi-
text, we forced this process with the encod- cates that the use of self during the encoding
ing task. In real life situations, it seems of adjectives produces as elaborate and inte-
grated a memory trace as has been found
likely that the self would be functional in a
using the present experimental paradigm.
number of situations involving feedback of
These data suggest the self is an extremely
personal data, such as conversations, expres-
important aspect of the processing of per-
sive behaviors, and attempts to assess per-
sonal information. In the realm of human
sonal impact on others. The kinds of situa-
information processing it is difficult to con-
tions that activate this schema or possible
ceive of an encoding device that carries more
individual differences in the frequency and potential for the rich embellishment of stim-
intensity of the involvement of self in data ulus input than does self-reference.
processing may prove to be very useful ex-
tensions of the present formulation.
Probably the main advantage of the pro- Reference Notes
cess-oriented view of information processing
1. Rogers, T. B., Rogers, P. J., & Kuiper, N. A.
underlying our approach to self is the oppor- Recognition memory for personal adjectives:
tunity to move toward less descriptive models Some evidence for self-reference as an aspect of
of social behavior. If we understand the memory. Unpublished manuscript, University of
processes and mechanisms underlying the Calgary, 1977.
processing of personal information, we will 2. Jackson, D. N. A threshold model for stylistic
responding. Paper presented at the meeting of
have a real opportunity to construct sub- the American Psychological Association, San
stantive models based on hard experimental Francisco, September 1968.
findings (see Sechrest, 1976). For example,
the finding that the self induces certain References
biases during the processing of personal data
(Cantor & Mischel, 1977; Markus, 1977) Allport, G. W. Personality: A psychological inter-
can be related to the cross-situational con- pretation. New York: Holt, 1937.
688 T. ROGERS, N. KUIPER, AND W. KIRKER

Bartlett, F. C. Remembering. Cambridge, England: Norms on 208 words typically used in their as-
Cambridge University Press, 1932. sessment. Canadian Journal of Psychology, 1972,
Bern, D. J., & Allen, A. On predicting some of the 26, 140-154.
people some of the time: The search for cross- Klein, K., & Saltz, E. Specifying the mechanisms
situational consistencies in behavior. Psychological in a levels-of-processing approach to memory.
Review, 1974, 81, SO6-S2O. Journal of Experimental Psychology: Human
Bertocci, P. A. The psychological self, the ego, and Learning and Memory, 1976, 2, 671-679.
personality. Psychological Review, 1945, 52, 91-99. Markus, H. Self-schemata and processing informa-
Cantor, N., & Mischel, W. Traits as prototypes: tion about the self. Journal of Personality and
Effects on recognition memory. Journal of Per- Social Psychology, 1977, 35, 63-78.
sonality and Social Psychology, 1977', 35, 38-48. Moscovitch, M., & Craik, F. I. M. Depth of
Cart wright, D. Self-consistency as a factor affecting processing, retrieval cues, and uniqueness of en-
immediate recall. Journal of Abnormal and Social coding as a factor in recall. Journal of Verbal
Psychology, 1956, 52, 212-219. Learning and Verbal Behavior, 1976, 15, 447-458.
Craik, F. I. M., & Lockart, R. S. Levels of process- Noble, C. E. An analysis of meaning. Psychological
ing: A framework for memory research. Journal Review, 1952, 59, 421-430.
of Verbal Learning and Verbal Behavior, 1972, Paivio, A. Neomentalism. Canadian Journal of Psy-
11, 671-684. chology, 1975, 29, 263-291.
Craik, F. I. M., & Tulving, E. Depth of processing Paivio, A., Yuille, J. C , & Rogers, T. B. Noun
and the retention of words in episodic memory. imagery and meaningfulness in free and serial
Journal of Experimental Psychology: General, recall. Journal of Experimental Psychology, 1969,
1975, 104, 268-294. 79, 509-514.
D'Agostino, P. R., O'Neill, B. J., & Paivio, A. Posner, M. I., & Keele, S. W. On the genesis of
Memory for pictures and words as a function of abstract ideas. Journal of Experimental Psychol-
levels of processing: Depth or dual coding? ogy, 1968, 77, 353-363.
Memory & Cognition, 1977, 5, 252-256. Posner, M. I , & Keele, S. W. Retention of abstract
Estes, W. K. Structural aspects of associative models ideas. Journal of Experimental Psychology, 1970,
for memory. In C. N. Cofer (Ed.), The structure 83, 304-308.
of human memory. San Francisco: Freeman, 1976. Rogers, T. B. An analysis of two central stages
Gergen, R. J. The concept of self. New York: underlying responding to personality items: The
Holt, Rinehart & Winston, 1971. self-referent decision and response selection. Jour-
Hastorf, A. H., Schneider, D. J., & Polefka, J. nal of Research in Personality, 1974, 8, 128-138.
Person perception. Don Mills, Ontario: Addison- Rogers, T. B. Self-reference in memory: Recogni-
Wesley, 1970. tion of personality items. Journal of Research in
Jackson, D. N. A manual for the Personality Re- Personality, in press.
search Form. Goshen, N.Y.: Research Psycholo- Schulman, A. I. Memory for words recently classi-
gists Press, 1967. fied. Memory & Cognition, 1974, 2, 47-52.
James, W. Principles of psychology. New York: Sechrest, L. Personality. In M. R. Rosenzweig &
Holt, 1890. L. W. Porter (Eds.), Annual review of psychol-
Jones, E. E., et al. (Eds.). Attribution: Perceiving ogy. Palo Alto: Annual Reviews, 1976.
the causes of success and failure. New York: Snygg, D., & Combs, A. W. Individual behavior.
General Learning Press, 1971. New York: Harper, 1949.
Jones, R. A., Sensenig, J., & Haley, J. V. Self- Walsh, D. A., & Jenkins, J. J. Effects of orienting
descriptions: Configurations of content and order tasks on free recall in incidental learning: "Diffi-
effects. Journal of Personality and Social Psy- culty," "effort," and "process" explanations. Jour-
chology, 1974, 30, 36-45. nal of Verbal Learning and Verbal Behavior, 1973,
Kelly, G. A. Psychology of personal constructs. 12, 481-488.
New York: Norton, 1955.
Kirby, D. M., & Gardner, R. C. Ethnic stereotypes: Received January 24, 1977 •