Product Familiarity and

Learning New Information

ERIC J. JOHNSON
J. EDWARD RUSSO*

Does product familiarity improve shoppers' ability to learn new product informa-
tion? We examine an earlier study which indicated that greater familiarity in-
creased learning during a new purchase decision. Our reanalysis confirms that
the effect depends strongly upon decision strategy. Familiarity facilitates learning
when consumers rate each alternative, but when consumers are instructed to
choose one alternative, an "inverted u" relationship between familiarity and learn-
ing results. Our new analyses also show that consumers familiar with the product
category demonstrate stronger brand organization for the new information.

Downloaded from http://jcr.oxfordjournals.org/ by guest on June 29, 2016

D uring the last decade it has become increasingly clear
that a decision maker's current knowledge of a topic
In a previous paper (Johnson and Russo 1981) we ex-
amined two plausible but conflicting hypotheses describing
affects the processing of new, topic-related information. In the relationship between learning and information acquisi-
consumer behavior, knowledge of a product class-or prod- tion. The first, which we term the "enrichment" hypoth-
uct familiarity-has been a feature of both traditional (Han- esis, suggests that existing knowledge facilitates the learn-
sen 1972; Howard 1977; Howard and Sheth 1969) and more ing of new information. A classic example is provided by
recent information processing theories of consumer choice the chess research of Chase and Simon (1973). In their
(Bettman 1979). Similarly, the impact of knowledge of a study, both chess masters and novices viewed actual chess
problem domain-or expertise-has been explored in many positions for five seconds. The chess masters' ability to
cognitive and social domains (see Chi, Glaser, and Rees recall these positions was superior to the novices' recall.
1981 for a review of the former, and Ostrom, Pryor, and With random patterns of chess pieces, however, the mas-
Simpson 1981; Fiske, Kinder, and Larter 1983 for examples ters' recall was no better than that of the novices. Thus
of the latter). Familiarity has been the focus of recent em- prior knowledge of the domain facilitated leaming-a "rich
pirical work in consumer research that examines informa- get richer" relationship which would generate data similar
tion acquisition (Bettman and Park 1980b), reactions to to the exponential curve in Figure A.
advertising (Anderson and Jolson 1980; Edell and Mitchell The second hypothesis suggests that prior knowledge has
1978; Marks and Olson 1981), and the choice of decision an "inverted u" effect, as shown in Figure A. Here, in
rules by consumers (Park 1976). The current paper has two contrast with the enrichment hypothesis, highly familiar
goals: consumers may search less than those who are moderately
~ To clarify the mechanisms underlying familiarity effects familiar. Bettman and Park (1980a) found such a pattern
in consumer choice. in consumers' acquisition of information about microwave
ovens, and Miyake and Norman (1979) found that the num-
~ To demonstrate the impact of familiarity upon consumers'
ability to search-and subsequently to learn-new infor- ber of questions asked about a new domain has an inverted
mation. u relationship with familiarity with similar domains. Al-
though both studies describe external information search,
this inverted u hypothesis can be extended to describe the
*Eric J. Johnson is Assistant Professor, Graduate School of Industrial amount of knowledge remembered after search, predicting
Administration, Carnegie-Mellon University, Pittsburgh, PA 15213. J. a curvilinear relationship between existing product knowl-
Edward Russo is Associate Professor, Graduate School of Business, Uni- edge and the amount of new information learned about a
versity of Chicago, Chicago IL 60637. The authors wish to thank Andrew product class.
Mitchell, Richard Staelin, and three anonymous reviewers for comments
on earlier drafts. This paper was prepared while the first author was a
How can we reconcile the different predictions made by
National Science Foundation Postdoctoral Fellow at the Department of the enrichment hypothesis and by the inverted u hypothesis?
Psychology, Stanford University. This work was also supported by Na- Familiarity with a product class could have several different
tional Science Foundation Grant #DAR76-81806 to the second author. A results, each of which might affect consumers' information
portion of this research was previously published in the 1981 Proceedings
processing skills in a different way. Familiarity gives ex-
of the Association for Consumer Research.
perienced consumers several advantages over consumers
542
© JOURNAL OF CONSUMER RESEARCH. Vol. 11 • June 1984
FAMILIARITY AND LEARNING 543

FIGURE A therefore remember-less of the externally available infor-

ALTERNATIVE THEORETICAL RELATIONSHIPS BETWEEN mation.
PRODUCT FAMILIARITY AND LEARNING To summarize, we have suggested three skills that de-
velop from increased familiarity with a product class, each
having consequences for the relationship between familiar-
ity and learning:
1. Familiar consumers possess superior knowledge of exist-
ing products, which should decrease search of extant al-
ternatives.
2. Familiar consumers possess a superior ability to encode
new information, which may increase search and learning
Amount of
for new alternatives.
Information
Recalled 3. Familiar consumers can pay attention to relevant infor-
mation and ignore irrelevant information: when faced with
Enrichment information that is irrelevant or redundant, such con-
Hypothesis sumers may use their knowledge of the product class to
ignore unimportant information, performing a more se-

Downloaded from http://jcr.oxfordjournals.org/ by guest on June 29, 2016

lective search of available information.
It is possible to isolate these types of familiarity effect
Product Familiarity
by considering decision tasks that make different demands
upon these three skills. For example, familiarity effects
arising from choices among new alternatives and existing
new to a product class. The first and most obvious is su- alternatives might differ: existing alternatives require the
perior knowledge of existing alternatives. Highly familiar use of familiar consumers' existing knowledge, while new
consumers-say, those with extensive knowledge of auto- alternatives make use of their superior encoding skills, and
mobiles-will be more likely to know specific facts about an increase in the information available for new alternatives
existing alternatives, such as the gas mileage of a Volks- might cause them to rely upon their superior ability to select
wagen Rabbit. As suggested by Bettman and Park (l980a), relevant information. Thus the impact of familiarity may
we expect knowledge to decrease search for highly familiar depend critically upon the relative importance of these three
consumers when they are considering existing alternatives. skills. Understanding the relationship between familiarity
A second distinct advantage of familiarity concerns and learning new information, in particular, requires a focus
search of new alternatives: more familiar consumers, like on the relative importance of superior encoding skills versus
the chess masters in Chase and Simon's studies, may de- superior information selection skills. We illustrate this by
velop knowledge about the plausible relationships among examining two similar tasks that we believe will produce
elements of a product class. Consumers who are familiar different relationships between familiarity and learning.
with autos, for example, may come to expect certain rela-
tionships between engine size, gas mileage, interior room,
and acceleration. This knowledge allows familiar con-
CHOICE VS. JUDGMENT
sumers to encode information about new alternatives more To clarify the role of task characteristics in mediating
efficiently and, as suggested by the enrichment hypothesis, familiarity effects, consider two tasks often associated with
causes an increase in learning. In addition to Chase and measurement of consumer preferences: a choice task
Simon's chess experiments, experts' superior ability to en- (choosing one alternative from a set), and a judgment task
code new information has been demonstrated in many do- (constructing an overall evaluation of an alternative). Many
mains, including baseball (Voss, Vesonder, and Spilich investigators (Bettman and Park 1980b; Johnson and Russo
1980), computer programming (Jeffries et al. 1981), and 1978; Wright and Barbour 1977) report evidence that the
electronics (Egan and Schwartz 1979). choice processes used by consumers are phased, combining
A third advantage of familiarity concerns the processing two decision procedures-such as elimination by aspects
of both novel and existing products. In many domains, a and additive utility-to make a choice. With such a rule,
key facet of expertise is the ability to select relevant infor- some alternatives are usually eliminated quickly on the ba-
mation while ignoring information irrelevant to the task at sis of one or two inferior values and not examined further.
hand (Larkin et al. 1980; Johnson 1983). More familiar Search using phased rules is usually quite selective, and
consumers may use their knowledge of the product class to consumers consequently have less knowledge of these elim-
limit their attention to information which is important to inated brands (Biehal and Chakravarti 1982; Johnson and
choice. At the same time, their superior encoding skills Russo, 1978).
may be offset by their superior ability to separate relevant Judgment, in contrast, requires a rating of each product
from irrelevant information: when the external environment on a scale of overall preference. This requires that overall
contains irrelevant information, experts may search-and judgments be made for each alternative. As in a linear com-
544 THE JOURNAL OF CONSUMER RESEARCH

pensatory strategy, we would expect the same amount of The present experiment used a brand x attribute display
information to be examined for each alternative (Payne of information abstracted from an actual advertisement.
1976). Since no alternative can be eliminated and the same This contained new information about subcompact cars that
attributes are examined for each alternative, the judgment had just become available at the beginning of the model
task limits the usefulness of information selection skills. year. Subjects who differed in their familiarity with auto-
These two tasks may well induce different strategies that mobiles were asked to evaluate the cars using only the
emphasize different skills of familiar consumers. When information presented. Afterwards, they were surprised
asked to judge the alternatives, subjects must construct an with instructions to recall this information. These recall
overall judgment for each alternative and consequently may protocols provided us with evidence of the role of prior
consider similar information for each alternative (Payne knowledge in learning (remembering) new product knowl-
1976). Here the superior encoding ability of more familiar edge. The recall protocols can also be analyzed to explore
consumers should dominate, producing a monotonic rela- the effect of familiarity on the content and organization of
tionship between familiarity and learning. The choice task this new knowledge.
instructions are less constraining, leaving the decision There were two limitations of our theoretical analysis and
maker free to select both strategy and information. Here experiment. First, we treated familiarity as synonymous
the information selection skills of the most experienced with knowledge about a domain (such as a class of prod-
consumers may dominate their superior encoding ability, ucts). For a consumer, this knowledge can come from many

Downloaded from http://jcr.oxfordjournals.org/ by guest on June 29, 2016

producing a decrease in external search and subsequent sources and is not necessarily correlated with usage. Al-
learning. Thus in the judgment task the superior encoding though learning can occur during the purchase and use of
skills of highly familiar consumers should prevail, produc- a product, other sources (such as advertising and word-of-
ing an enrichment effect. In choice, a highly familiar con- mouth) can create familiarity without direct experience. In
sumer's ability to select information should come into play, this study, therefore, we described familiarity as knowledge
causing a decrease in learning new information that con- of the product class and not necessarily as usage experience.
forms to the inverted u hypothesis. Second, there are many ways to learn about a product
Incidentally, this distinction between choice tasks and class, and no one study can examine them all. We limited
judgment tasks is also found in the behavioral decision lit- ourselves to one type of learning: the acquisition of new
erature. As Einhorn and Hogarth (1981) point out, com- knowledge that occurs during purchase decisions. This ap-
pensatory processes are not necessary for psychological proach has two advantages: (1) there is evidence that im-
models of choice. The process-tracing evidence both in be- portant parts of consumers' product knowledge come from
havioral decision theory (Payne 1976) and in consumer be- this source (Wright 1975; Johnson and Russo 1978), and
havior (Bettman and Zins 1977; Lussier and Olshavsky (2) the nature of information learned during purchase de-
1979) suggests that the use of compensatory strategies is cisions gives us some insight into the relationship between
limited to situations with few alternatives. However, other familiarity and the processes used to make the decision.
research suggests that unfamiliar consumers may make
greater use of additive compensatory-decision rules (Park
1976). METHOD
Subjects
OVERVIEW
The participants were 55 Master's students in business
To summarize, this research concerns prior knowledge who completed the task as part of a classroom demonstra-
of a product class and its effect upon consumers' learning tion during the first meeting of a consumer behavior course.
of new information. The literature suggests two alternative Although we selected the product category-new subcom-
hypotheses: pact cars-because it was of substantial interest to this
1. An enrichment hypothesis, which suggests that greater group, the results from this convenience sample may not
prior knowledge leads to more extensive leaming of new generalize to other populations, especially to those with
information. different distributions of product familiarity.
2. The inverted u hypothesis, which suggests that higher
levels of familiarity result in reduced search and less Stimulus
learning.
An edited portion of a newspaper advertisement placed
We suggest that the first hypothesis may reflect familiar by General Motors' Oldsmobile Division served as the
consumers' superior encoding, while the latter reflects brand x attribute matrix. It compared eight small imported
highly familiar consumers' ability to eliminate useless or cars on 11 attributes. Further details are available in John-
redundant information. We explore these two hypotheses son and Russo (1981).
across two tasks: a choice task that allows subjects consid- Using actual products presents a greater possibility that
erable freedom in their selection of decision processes, and the recall protocols might be contaminated by intrusions
a judgment instruction designed to promote compensatory from prior knowledge. This brand x attribute matrix, how-
processing. ever, presented obscure variants of relatively common au-
FAMILIARITY AND LEARNING 545

tos, specifying nonstandard engines, transmissions, and so RESULTS

on. This minimized the problem of separating knowledge
acquired prior to the experiment from our dependent mea- Familiarity
sures of knowledge acquired through making a decision.
Moreover, pretesting showed that even knowledgeable con- We can consider the possible relationship between fa-
sumers had difficulty generating the matrix if they had not miliarity and learning by examining the total amount of
seen it, and that they were unable to construct much of it knowledge recalled as a function of familiarity. Previous
accurately. The advantage of this procedure is that it al- analyses (Johnson and Russo 1981) showed the expected
lowed us to present even expert consumers with information interaction between familiarity and the evaluation task. For
that they did not know, while maintaining the meaningful judgment, the enrichment effect was large and consistent:
patterns present among the attributes in the product class. the mean number of statements recalled was 11.6 for the
low familiarity group, 14.4 for those of moderate familiar-
Procedure ity, and 28.6 for those high in product familiarity. The
choice condition, in contrast, revealed a large and reliable
Subjects were run in a single group and assigned ran- inverted u relationship: consumers who had been asked to
domly to task conditions. Each group received one of the make choices recalled an average of 12.3, 19.3, and 15.7
two sets of task instructions, which asked them to rate each statements in the low, medium, and high familiarity groups,
alternative (the judgment condition), or to choose one from respectively. An ANOV A confirmed the significance of the

Downloaded from http://jcr.oxfordjournals.org/ by guest on June 29, 2016

the set (the choice condition). Further details are provided interaction, and a priori contrasts demonstrated that both
in Johnson and Russo (1981). the linear effect in judgment and the inverted u effect in
Upon completion of their tasks (typically within 5 min- choice were statistically significant.
utes), both groups completed a "demographic question- This same pattern is shown by a new structural equation
naire" that prevented retention of the product information analysis that has two additional advantages: (I) it uses the
in short-term memory and provided a self-rating of product multiple measures of familiarity collected in this study, and
familiarity. Subjects noted their previous knowledge of au- (2) it does not depend upon arbitrary cutoffs to determine
tomobiles, compared to the rest of the popUlation, on a the level of familiarity. In this analysis, familiarity is mea-
five-point scale and reported their usage experience in the sured by the self-report scale described above and by two
product class. 1 other self-report measures of experience: the reported num-
Afterwards, subjects turned the page and read the instruc- ber of cars in this product class owned by the respondent,
tions for an unexpected written recall task. They were asked and the number that s/he had ridden in. In addition to the
to "not only include the information you were given, but latent construct, familiarity, the structural model included
also any other judgments or comparisons you can remem- the effect of task, which we assumed was measured without
ber." These instructions encouraged complete recall, in- error. Since the hypothesized effect of familiarity in the
cluding tentative judgments about the products made during choice condition was nonlinear, we need to operationalize
the course of the decision. When they had finished, subjects some nonlinear form of the unobservable construct of fa-
were asked to report which car they preferred and any past miliarity. An obvious candidate is the square of the con-
experience they had with the eight automobiles used for the struct, which could be operationalized by imposing con-
task. straints across the measurement models for familiarity,
operationalizing the inverted u effect. However, the avail-
Analysis able algorithms commonly used for the estimation of struc-
tural equation models such as L1SREL do not allow us to
We measured familiarity with the product category using impose the necessary constraints. We therefore calculated
subjects' self-reports on the scale described above. Self- the nonlinear effect of familiarity externally to the estima-
ratings below 3 were classified in the low familiarity group, tion of the structural equation. We imposed the necessary
those between 3 and 4 in the moderate familiarity group, constraints by using the coefficients calculated for the linear
and those above 4 in the high familiarity group. These component of familiarity, which were then substituted in
integer values used to create the familiarity groups resulted the calculation of the nonlinear effect. We iterated this es-
in unequal cell sizes (16, 27, and 12 subjects in the low, timation process until stable sets of coefficients emerged. 2
medium, and high familiarity conditions, respectively). The Because we expected the nonlinear, inverted u effect of
two task instructions-judgment versus choice-along with familiarity only in the choice condition, we set the value
the three levels of familiarity created a 2 x 3 factorial of the nonlinear effect to zero for the observations from the
design with unequal cell sizes. judgment group.

'Validity of the measure is a problem if the self-report actually measures 2This process can yield liberal significance tests, however, because it
some other construct that would explain these results. Although this mea- assumes that there is no error in the measurement of the nonlinear effect
sure has apparent face validity, there is one plausible alternative construct: of familiarity. Estimation of a model using the appropriate constraints with
motivation or interest in the product class. We address this alternative in a nonlinear estimation procedure (HOTZTRAN, Avery and Hotz 1983)
a later section. yielded similar results.
546 THE JOURNAL OF CONSUMER RESEARCH

FIGURE B
STRUCTURAL EQUATION MODEL OF THE RELATIONSHIP BETWEEN FAMILIARITY AND LEARNING

2.28*

Downloaded from http://jcr.oxfordjournals.org/ by guest on June 29, 2016

-1.25*

(} = \.63* 109.4*
8 -0.36 0.33 1.9 0.29*
-0.22 -0.53 0.64 -143.3 * -3.76 212.6
NOTE: x, = Number of cars ridden F = Familiarity
x2 = Number of cars owned F2 = Inverted u familiarity effect
x3 = Self-report of familiarity • = Effect significant at p < 0.05

Figure B displays the model and the estimated coeffi- fect shows considerable lack of fit: a o.} = 0.360. The ad-
cents_ The coefficients conform to the results of the simple dition of the nonlinear effect of familiarity shows that the
ANOV A, showing a significant positive linear relationship increase in fit between M} and M 2 is significant: X2 (l df)
between familiarity and recall and a significant negative = 43.78, p < 0.001. Thus the structural equation analysis
relationship between the square of the familiarity construct tells a story that is similar to the ANOVA, suggesting that
and recall. The coefficient for task is not significant. One the enrichment hypothesis describes information recalled
of the three measures of familiarity-the number of cars after a judgment but not the recall of consumers who have
driven-is scaled to a unit variance. The remaining two made choices.
measures show a strong relationship to the underlying con-
struct of familiarity.
Overall, the model provides a reasonable fit to the data: Choice versus Judgment
X2 (4 df) = 3.22, p > 0.50. To further explore competing
models of the data we compared: Why does familiarity have such a different effect upon
judgment and choice tasks? We suspect that the answer lies
1. Mo, a null model that estimated all variances and covari- in the different cognitive processes underlying choice and
ances between the observed variables but no measurement judgment. Specifically, we suspect that a choice task allows
or structural parameters, the use of phased decision rules that eliminate alternatives
2. M" a model that estimated all but one parameter of the and thus cause more to be known about the brand that is
model shown in Figure B, setting the nonlinear effect of eventually chosen. In a judgment task, such elimination is
familiarity to zero, and problematic, so we expect search and subsequent recall to
3. M 2 , the full model displayed in Figure B.
be about equal for all alternatives.
Although we did not collect data examining this issue in
Since these models are nested, we can calculate both the main experiment, there is process-tracing evidence con-
incremental indices of fit and tests of the significance of the sistent with this speculation (Johnson and Meyer 1984).
improvements due to the added parameters (Bentler and Specifically, this evidence indicates that the verbal proto-
Bonett 1980). The normed index of fit-a o2-is quite rea- cols generated during choice differ from those generated
sonable (0.957), while the model lacking the nonlinear ef- during judgments in three ways:
FAMILIARITY AND LEARNING 547

1. Choice protocols contain relatively more statements elim- iarity consumers, 0.32 and 0.27 of all statements referred
inating alternatives. to the chosen brand, while 0.54 of the high familiarity
2. Choice protocols contain relatively fewer statements that group's statements referred to the car that they had chosen.
evaluate alternatives. An ANOV A confirmed the task x level interaction: F
(2,49) = 5.82, p < 0.01. A priori comparisons confirmed
3. Choice protocols demonstrate a one-sidedness effect- the reliability of increases in the choice condition, F (1,49)
that is, they show that much more attention is paid to the = 11.20, P < 0.005, and found no effect in the judgment
alternative eventually chosen.
condition. This supports the hypothesis that consumers in
We conducted a small pilot study with the current stimuli the choice condition use decision rules that eliminate alter-
to examine the possibility that similar differences in strategy natives. It also suggests that the most familiar consumers
may underlie the familiarity x task interaction. Ten un- may use choice rules that are particularly one-sided. In
dergraduates generated concurrent verbal reports: half per- contrast, the judgment condition provides little evidence of
formed the choice task from the current experiment, and selective memory.
half performed the judgment task. An analysis of these
reports, using a coding scheme similar to that employed by Organization
Johnson and Meyer (1984), shows similar differences be-
While we have concentrated on the amount of knowledge
tween choice and judgment tasks:
retained by consumers, recall protocols also allow us to

Downloaded from http://jcr.oxfordjournals.org/ by guest on June 29, 2016

1. The choice protocols contained elimination statements (8 make tentative inferences concerning the organization of
percent of all statements), while the judgment protocols this knowledge in memory. This organization can be iden-
contained none. tified, in part, by the order of statements in the recall pro-
2. The judgment protocols consisted largely of statements tocols. If memory is based upon brand organization, we
coded as evaluations (44 percent), while these statements expect all the facts about a brand to be clustered together
represented only 7 percent of the protocols generated in in recall. Attribute organization occurs when information
the choice task. about many brands is clustered around one attribute. An
3. The choice condition shows a similar one-sidedness of index of organization can be constructed by counting all
search: 27 percent of all statements refer to the most pre- transitions between statements and dividing them into three
ferred alternative, whereas the judgment condition sub- groups: those that share the same brand, those that share
jects mentioned the most preferred alternative in only 9 the same attribute, and those that contain neither the same
percent of their statements. brand nor the same attribute. These are called brand, at-
tribute, or neutral transitions, respectively (Johnson and
Although the groups are small, all three differences are Russo 1978). The proportion of brand transitions for each
significant: Mann-Whitney U = 3, 0, 2, p < 0.05, 0.01, cell is listed in the Table.
and 0.05, respectively. Thus this independent pretest in- These data clearly show a large effect of familiarity upon
dicates that the two sets of task instructions used in this organization: as familiarity increases, acquired knowledge
experiment may well have produced markedly different sets shows increasing brand-based organization. This effect is
of cognitive processes: choice appears to be characterized significant-F (2,49) = 7.84, P < O.OOI-while the effect
by elimination and "one-sided" search, while judgment of task and the interaction are not-F (l,49) = 2.16 and
appears to imply more evaluation and a more balanced pat- F (2,49) = 1.79, p > 0.15 for both. The proportion of
tern of search. attribute-based transitions shows a similar decrease. In sum,
These apparent processing differences help explain dif- the organization data present a clear picture: more famil-
ferences in the information recalled in each task condition iarity results in more brand organization.
which we had previously observed in Johnson and Russo
(1981) . We examined the recall protocols for evidence of
one-sidedness by counting the number of statements that
Controls for Prior Knowledge
referred to the automobile most preferred by each subject. Our analysis rests on the assumption that these recall
We found that 35 percent of all statements in the choice protocols contain relatively few facts known before the ex-
task referred to the chosen automobile, while only 18 per- periment. This neglects the possibility that because they
cent of the statements in the judgment task mentioned the have more prior knowledge of ~hese autos, more familiar
most preferred car. Both values are significantly greater consumers may include this knowledge in their recall, art-
than the 12.5 percent expected by chance (p < 0.001). ificially raising their total for newly acquired knowledge.
We also examined these proportions in each task and This explanation cannot account for the more interesting
level of familiarity. The judgment condition showed no results-i.e., the observed interaction between familiarity
statistically significant effect of familiarity: the proportion and task instructions and the decreased recall of experienced
of statements referring to the chosen alternative was 0.13, subjects in the choice conditions-but extensive recall of
0.23, and 0.12 for the low, medium, and high levels of prior knowledge would affect the interpretation of these
familiarity, respectively. However, the choice task showed results.
an increase in one-sidedness that was particularly large for Although we cannot conclusively show that no prior
highly familiar consumers: for the low and medium famil- knowledge was recalled, we examined these protocols for
548 THE JOURNAL OF CONSUMER RESEARCH

TABLE determine whether the differences in recall we observed are

PROPORTION BRAND TRANSITIONS due to encoding or to retrieval. Like most researchers who
have examined recall as a function of familiarity or exper-
Low Moderate High tise, we assume that differences in recall are due to differ-
familiarity familiarity familiarity Mean ences in encoding. However, additional research would be
necessary to study differences in experts' and novices' strat-
Choice .26 .54 .58 .46 egies for recall in similar tasks.
Judgment .21 .31 .69 .36
Mean .24 .41 .63 .27
DISCUSSION
Our three levels of familiarity can be seen as three snap-
inaccurate recall, which could have been due in part to shots portraying stages in the development of an "expert"
recalled prior knowledge, since the autos were often ob- consumer-one knowledgeable about the product category.
scure variants of a manufacturer's line and the matrix con- We can trace this development by comparing the changes
tained some attributes that were not well publicized (EPA across the conditions. Concentrating on the judgment task,
estimates of trunk capacity, for example). In addition, none we see that prior knowledge enhances a consumer's ability

Downloaded from http://jcr.oxfordjournals.org/ by guest on June 29, 2016

of the recall protocols referred to brands or attributes that to encode and remember new information. In addition to
were not mentioned in the original matrix. Pretesting con- this increased encoding ability, more familiar consumers
firmed that even knowledgeable consumers had great dif- develop more brand organization for new knowledge. This
ficulty completing most of the array if they had not seen is consistent with a number of views of learning which
the matrix beforehand. 1 his suggests that reconstruction suggest that acquiring new knowledge also entails devel-
from prior knowledge should not be as accurate as knowl- oping unitized knowledge structures (see Hayes-Roth 1977
edge acquired from the display. for a particularly appropriate example). Whether we call
A rater who was blind to our hypotheses scored the brand these units frames, scripts, schemata, or simply chunks
attribute values in the protocols for accuracy. A strict cri- does not matter. The evidence clearly indicates the devel-
terion for accuracy was that values of all verbal descriptions opment of an integrated unit of storage and suggests that
be recalled exactly and that all numerical attributes be re- this unit is the brand.
called within 2 percent of their correct value. Analysis using The results from the choice task demonstrate that some-
this strict criterion showed that: thing else is also happening. Evidently, experienced con-
sumers use their knowledge of the product class to limit
• The proportion of inaccurate recall (0.37) did not vary their search. In the real world outside of this experiment,
with either familiarity or task condition (p > 0.20).
this decrease could be due to prior knowledge of informa-
• Inclusion of the number of inaccurate statements as a co- tion presented in the environment (Bettman and Park 1980).
variate did not substantially change the significance of the In this experiment, however, the knowledge presented was
results in the reported analysis of variance.
not available even to experienced consumers, indicating
These results strongly suggest that the knowledge recalled that presence of the knowledge in memory was not respon-
by our subjects was acquired from the display and did not sible for reduced search. Rather, it appears that, along with
represent prior knowledge of the presented values. their increase in encoding ability, experienced consumers
An alternative explanation for these results concerns mo- develop knowledge of efficient decision procedures. For
tivational differences among the groups. The pattern of re- example, experienced consumers would ignore the attribute
sults for anyone group is easily explained from a motiva- "cruising range," since they realize that it is simply the
tional perspective. For example, more interested and product of fuel capacity and estimated miles per gallon.
motivated subjects might learn more information about the This procedural knowledge appears to be a major advantage
product class, yielding the data consistent with the enrich- that experienced consumers bring to decision tasks. 3
ment hypothesis seen in the judgment task. It is not clear,
however, why this would cause a decrease in search in the
choice task condition. We suggest that the decrease comes IMPLICATIONS
not from disinterest in the product class, but from knowl-
edge that further search would not lead to a better choice. These results really have two separate stories to tell: one
Any alternative explanation of these effects needs to ac- about differences among decision tasks, and the other about
count for: (1) the increase in recall with familiarity in the the effect of familiarity. Each has implications for broader
judgment task, (2) the equivalent inverted u pattern in the
choice task, and (3) the one-sidedness observed in choice 3Interestingly, Punj and Staelin (1983) did not find an inverted u rela-
but not in judgment. These phenomena suggest that the tionship using self-reports of knowledge made by actual car purchasers.
This could be due to the limited range of familiarity among their subjects,
information processing explanation may be a more parsi- all of whom had purchased cars within six months prior to the study. Thus
monious account for these results, but that further research more research reconciling the results of field and laboratory studies of
seems warranted. Further research would also be needed to information search seems warranted.
FAMILIARITY AND LEARNING 549

areas within consumer behavior. We will discuss the im- technical ad is most effective for those with no experience.
plications of task differences for the estimation of choice One special class of experienced consumers consists of in-
models and then tum to familiarity, discussing the impli- dustrial purchasers and professional buyers, who must be
cations of the current findings for information provision. considered "expert" consumers. Our data suggest that,
barring time pressure, highly technical appeals containing
much information might be effective for this unique seg-
Choice versus Judgment
ment.
Our data suggest that choosing one alternative from a set Second, our results emphasize the importance of learning
can invoke different psychological processes than judging procedures for making decisions. Consumer Reports is
alternatives, which are presumably evaluated one at a time. often cited as the classic instance of a brand x attribute
In choice, the memory data are consistent with the use of display. The text accompanying the display is often not
elimination-based strategies, while judgment seems to re- mentioned, but is an excellent example of the kind of pro-
sult in different, more compensatory processing. However, cedural knowledge available to experienced consumers: it
most models of consumer preferences appear to equate the identifies important attributes, describes the range of values
two. for attributes, and suggests reasonable levels for cutoffs.
For example, decision decomposition techniques, such The impact of this information in helping naive consumers
as conjoint analysis, estimate attribute importance by asking make better decisions may be as great as the effect of the
consumers to judge or rank alternative products. The

Downloaded from http://jcr.oxfordjournals.org/ by guest on June 29, 2016

brand x attribute display. This raises the possibility of
weights derived from these one-at-a-time judgments are procedural education, in which similar descriptions act as
then used to predict consumers' selections in a real-world effective supplements to the brand x attribute matrix. Pro-
choice task. The picture of attribute usage that emerges in cedural education is a technique that is open to advertisers
this experiment is quite different for these two tasks. as well as to those who wish to improve decisionmaking
This raises several interesting issues. What happens, for though public policy.
example, when choice strategies eliminate alternatives? This suggests a third and final implication. Since publi-
According to our data, elimination does not seem com- cation of the work on information overload by Jacoby and
monplace in one-at-a-time judgments. This implies that colleagues (Jacoby, Speller, and Berning 1974; Jacoby,
weights estimated using judgments may not have a clear Speller, and Kohn 1974), a major controversy in consumer
relationship to attribute usage in choice. Although this does behavior research has concerned the amount of information
not argue against the usefulness of decomposition models, provided to consumers. When faced with a large number
which have been shown to be useful predictors of choice, of attributes and limited processing capacity, both experi-
these results may help delineate their limits. Recent results enced and naive consumers consider a subset of the avail-
reported by Huber and Czajka (1982) illustrate the point. able information (Bettman and Kakkar 1977; Payne 1976).
They showed that two different tasks-binary choice and Our results suggest that experienced consumers should be
one-at-a-time judgments-yielded slightly different rank better able to select attributes that are predictive of product
orderings of attribute importance. Since elimination-based performance, which should, in tum, result in better deci-
strategies appear to be more common in choice tasks with sions. The implication for the provision of information to
greater than two alternatives (Payne 1976), it may be that naIve consumers is clear: presentations should be limited
disagreement between tasks is a more important problem to the attributes most relevant to preference judgments.
with larger choice sets. There are other ways of measuring Additional information, particularly when it is nondiagnos-
preferences, such as ranking, and a process analysis of the tic of product performance, could cause deterioration in the
influence of task upon decomposition models is published quality of choice.
elsewhere in this issue (Johnson and Meyer 1984).
[Received August 1982. Revised November 1983.)
Familiarity REFERENCES
The other story to be told concerns the effect of famil- Anderson, Rolph E. and Marvin A. Jolson (1980), "Technical
iarity. This research, along with related findings, has sev- Wording in Advertising Implications for Market Segmenta-
eral implications for both public and private providers of tion," Journal of Marketing, 44 (1),57-66.
product information. First, our findings suggest that famil- Avery, Robert B. and V. Joseph Hotz (1983), HOTZTRAN Users
iarity provides a useful segmentation technique. Com- Manual, technical report, Carnegie-MeHon University, Pitts-
munication should match the technical complexity of its burgh, PA, 15232.
Biehal, Gabriel J. and Dipankar Chakravarti (1982), "Informa-
intended audience. Recent research on responses to adver-
tion-Presentation Format and Learning Goals as Determinants
tising confirms this: Edell and Mitchell (1978) show that of Consumers' Memory Retrieval and Choice Processes,"
highly familiar consumers report more cognitive responses Journal of Consumer Research, 8 (March), 431-441.
when presented with technical advertising, and Anderson Bentler, Peter M. and Douglas G. Bonett (1980), "Significance
and Jolson (1980) demonstrate that technical ads create the Tests and Goodness of Fit in the Analysis of Covariance
greatest increase in purchase intentions for purchasers who Structures," Psychological Bulletin, 88 (November),
have considerable experience with the product, while a non- 588-606.
550 THE JOURNAL OF CONSUMER RESEARCH

Bettman, James R. (1979), An Information Processing Theory of Jeffries, Robin, Althea A. Turner, Peter G. Polson, and Michael
Consumer Choice, Reading, MA: Addison-Wesley. E. Atwood (1981), "Processes Involved in Designing Soft-
- - - and Pradeep Kakkar (1977), "Effects of Information Pre- ware," in Cognitive Skills and Their Acquisition, ed. J. R.
sentation Format on Consumer Information Acquisition Strat- Anderson, Hillsdale, NJ: Lawrence Erlbaum.
egies," Journal of Consumer Research, 3 (March), 233- Johnson, Eric J. (forthcoming), "Expertise in Decision Making:
240. Process and Performance," in The Nature of Expertise, eds.
- - - and C. Whan Park (1980a), "Implications of a Construc- Micheline Chi et aI., in preparation.
tive View of Choice for Analysis of Protocol Data: A Coding - - - and Robert J. Meyer (1984), "Compensatory Choice
Scheme for Elements of Choice Processes," in Advances in Models of Noncompensatory Choice Processes: The Effect
Consumer Research, Vol. 7, ed. Jerry C. Olson, Ann Arbor, of Varying Context," Journal of Consumer Research, 11
MI: Association for Consumer Research. (June), 528-541.
- - - and C. Whan Park (1980b), "Effects of Prior Knowledge - - - and J. Edward Russo (1978), "What is Remembered after
and Experience and Phase of the Choice Process on Con- a Purchase Decision," working paper, Graduate School of
sumer Decision Processes: A Protocol Analysis," Journal of Business, University of Chicago, Chicago, IL.
Consumer Research, 7 (December), 234-248. - - - and J. Edward Russo (1981), "Product Familiarity and
- - - and Michael Zins (1977), "Constructive Processes in Learning New Information," in Advances in Consumer Re-
Consumer Choice," Journal of Consumer Research, 4 (Sep- search, Vol. 8, ed. Kent B. Monroe, Ann Arbor, Ml: As-
tember), 75-85. sociation for Consumer Research.
Chase, William G. and Herbert A. Simon (1973), "Perception in Larkin, Jill H., John McDermott, Dorothea P. Simon, and Herbert

Downloaded from http://jcr.oxfordjournals.org/ by guest on June 29, 2016

Chess," Cognitive Psychology, 4 (January), 55-81. A. Simon (1980), "Models of Competence in Solving Phys-
Chi, Micheline T. H., Robert Glaser, and Ernest Rees (1981), ics Problems," Cognitive Science, 4 (4),317-345.
"Expertise in Problem Solving," in Advances in the Psy- Lussier, Dennis A. and Richard W. Olshavsky (1979), "Task
chology of Human Intelligence, ed. R. Sternberg, Hillsdale, Complexity and Contingent Processing in Brand Choice,"
NJ: Lawrence Erlbaum. Journal of Consumer Research, 6 (September), 93-100.
Edell, Julie A. and Andrew A. Mitchell (1978), "An Information Marks, Larry J. and Jerry C. Olson (1981), "Toward a Cognitive
Processing Approach to Cognitive Responses," in Research Structure Conceptualization of Product Familiarity," in Ad-
Frontiers in Marketing Dialogues and Directions, ed. S. C. vances in Consumer Research, Vol. 8, ed. Kent B. Monroe,
Jain, Chicago: American Marketing Association. Ann Arbor, MI: Association for Consumer Research.
Egan, D. and B. Schwartz (1979), "Chunking in Recall of Sym- Miyake, Naomi and Donald Norman (1979), "To Ask a Question,
bolic Drawings," Memory and Cognition, 7 (March), One Must Know Enough to Know What Is Not Known,"
149-158. Journal of Verbal Learning and Verbal Behavior, 18 (June),
Einhorn, Hillel J. and Robyn M. Hogarth (1981), "Behavioral 357-364.
Decision Theory: Processes of Judgment and Choice," An- Ostrom, Thomas, John Pryor, and David Simpson (1981), "The
nual Review of Psychology, 32, 52-88. Organization of Social Information," in Social Cognition:
Fiske, Susan, Donald Kinder, and W. Michael Larter (1983), The Ontario Symposium, eds. E. T. Higgins et aI., Hillsdale,
"The Novice and the Expert: Knowledge-Based Strategies NJ: Lawrence Erlbaum.
in Political Cognition," Journal of Experimental Social Psy- Park, C. Whan (1976), "The Effect of Individual- and Situation-
chology, 19 (4),381-400. Related Factors on Consumer Selection of Judgmental
Hansen, Flemming (1972), Consumer Choice Behavior, New Models," Journal of Marketing Research, 13 (May),
York: Free Press. 144-151.
Hayes-Roth, Barbara (1977), "Evolution of Cognitive Structure Payne, John W. (1976), "Task Complexity and Contingent Pro-
and Processes," Psychological Review, 84 (May), 260-278. cessing in Decision Making: An Information Search and Pro-
Howard, John A. (1977), A Consumer Behavior: Application of tocol Analysis," Organizational Behavior and Human Per-
Theory, New York: McGraw-Hill. formance, 16 (August), 366-387.
- - - and Jagdish N. Sheth (1969), The Theory of Buyer Be- Punj, Girish N. and Richard Staelin (1983), "A Model of Con-
havior, New York: John Wiley. sumer Information Search Behavior for New Automobiles,"
Huber, Joel and Andrew L. Czajka (1982), "Modeling Buy/No Journal of Consumer Research, 9 (March), 366-380.
Buy Decisions: A Comparison of Two Methods," in Ad- Voss, James F., Gregg T. Vesonder, and George 1. Spilich
vances in Consumer Research, Vol. 9, ed. Andrew A. Mitch- (1980), "Text Generation and Recall by High-Knowledge
ell, Ann Arbor, MI: Association for Consumer Research. and Low-Knowledge Individuals," Journal of Verbal Learn-
Jacoby, Jacob, Donald E. Speller, and Carol Kohn (1974), ing and Verbal Behavior, 19 (December), 651-667.
"Brand Choice Behavior as a Function of Information Wright, Peter L. (1975), "Consumer Choice Strategies: Simpli-
Load," Journal of Marketing Research, 11 (February), fying vs. Optimizing," Journal of Marketing Research, 11
63-69. (February), 60-67.
- - - , Donald E. Speller, and Carol A. Kohn Berning (1974), - - - and Frederic Barbour (1977), "Phased Decision Strate-
"Brand Choice Behavior as a Function of Information Load: gies: Sequels to an Initial Screening," in North Holland/
Replication and Extension," Journal of Consumer Research, T1MS Studies in the Management Science, eds. Martin K.
1 (June), 33-42. Starr and Milan Zeleny, Amsterdam: North Holland.