The Effects of Part-List Cuing on Attribute Recall: Problem Framing At the Point of Retrieval

ABSTRACT - Previous research has demonstrated that advertising may frame a decision at the time of product learning by forcing the consumer to focus on particular product attributes to the exclusion of others. The present research demonstrates that problem framing may also occur at the time of retrieval. In two experiments it was shown that information provided at the time of retrieval can inhibit recall, and therefore consideration, of well-known attributes of a familiar product.

Citation:

Joseph W. Alba and Amitava Chattopadhyay (1985) ,"The Effects of Part-List Cuing on Attribute Recall: Problem Framing At the Point of Retrieval", in NA - Advances in Consumer Research Volume 12, eds. Elizabeth C. Hirschman and Moris B. Holbrook, Provo, UT : Association for Consumer Research, Pages: 410-413.

Advances in Consumer Research Volume 12, 1985      Pages 410-413

THE EFFECTS OF PART-LIST CUING ON ATTRIBUTE RECALL: PROBLEM FRAMING AT THE POINT OF RETRIEVAL

Joseph W. Alba, University of Florida

Amitava Chattopadhyay, University of Florida

ABSTRACT -

Previous research has demonstrated that advertising may frame a decision at the time of product learning by forcing the consumer to focus on particular product attributes to the exclusion of others. The present research demonstrates that problem framing may also occur at the time of retrieval. In two experiments it was shown that information provided at the time of retrieval can inhibit recall, and therefore consideration, of well-known attributes of a familiar product.

INTRODUCTION

It is widely known that an advertiser's goal may range from the prosaic (e.g., creating brand familiarity) to the ambitious (e.g., altering behavior). If, as some have suggested, a primary goal is to influence a consumer's attitudinal structure, then a variety of strategies are available for attaining it (Boyd, Ray, and Strong 1972). One common strategy is to change the beliefs consumers hold about a product's and/or its competitors' attributes. A second strategy is to add attributes to the product and 'then convince consumers that these new attributes should be considered during the decision process. Finally, an advertiser may attempt to change the salience of particular attributes, and thereby effectively change the importance weights consumers assign to them. These latter two strategies are examples of what has become known as problem framing.

It has long been believe I that although advertising may fall short of outright persuasion, it may subtly alter perception of a product by increasing the salience of particular attributes (Krugman 1965). Wright and Rip (1980) have noted, however, that advertisers may be subtle or overt in the methods they use to alter the attributes consumers consider and/or the weights they assign to them. For example, by frequently and consistently describing attributes of a product class in a particular manner, advertisers may lead consumers to consider primarily those attributes during evaluation. On the other hand, an advertiser may overtly tell the consumer to "frame it my way" by explicitly stating that certain attributes are more important than others. Evidence exists to show that attribute recall and brand attitude can each be influenced by the amount of emphasis placed on a product's attributes (Gartner 1983: Wright and Rip 1980).

Not surprisingly, the research on problem framing has focused directly on the influence that the message has on the attributes that are described or emphasized. Consequently, when framing effects are found, one explanation advanced to account for them is that emphasized attributes receive more attention, thereby making them more salient (important) and/or more memorable to the decision maker. The important point is that problem framing, as previously studied, affects the amount of attention the consumer pays to certain attributes during the encoding, or learning, of product information.

A complementary way of looking at problem framing is to focus on processes that occur during retrieval. That is, a decision may be framed after all relevant information has been learned. If so, framing can occur independently of attentional factors operating during encoding.

One way in which a problem may be framed during retrieval is through the use of part-list cuing (cf. Lynch and Srull 1982). Research has amply demonstrated that cues presented at time of retrieval may enhance recall, especially if they provide access to categories of information not previously recalled (Tulving and Pearlstone 1966; Tulving and Psotka 1971). However, if the set of to-be-remembered items is relatively homogenous (i.e., do not come from different categories), a subset of those items presented as cues may actually inhibit recall of the remaining items (Roediger 1974). Although this phenomenon has been studied almost exclusively within psychology using very neutral stimuli, the implications for consumer behavior are obvious. For examples if a product possesses several attributes, presentation of a subset of them at the point of attitude formation may inhibit recall, and therefore consideration, of the remaining attributes. This situation arises quite frequently in the case of advertising. An advertisement typically contains information only about attributes that reflect favorably on the brand. Similarly, packages and displays emphasize only favorable information. Although this serves to increase the salience of favorable attributes, it may also inadvertently inhibit recall of the brand's unattractive attributes. In so doing, it frames the problem.

Note that this method of problem framing represents a subtle shift in perspective. Part-list cuing works not by directly persuading the consumer to place higher weights on the attributes emphasized at the time of learning, but by actively inhibiting consideration of familiar attributes not cued at the time of recall (cf. Gardner 1983).

The most popular explanation of part-list cuing inhibition is provided by Rundus (1973). The aspects of his model that are important for present purposes can be described briefly. First, it is assumed that recall of a category of information is guided by a superordinate cue, usually the category name. Search proceeds from the category name to the individual category items connected to it. For example, the product name "automobile" could be used as a cue to retrieve all attributes associated with the product. It is further assumed that retrieval of items from memory is characterized by sampling with replacement. Thus, once an item (attribute) from a category is recalled, it may be sampled again during subsequent retrieval attempts. Consequently, the more items from the pool that are recalled, the lower is the probability that a recall attempt will produce a previously unrecalled item. As recall proceeds, production of new items becomes more difficult. In the case of part-list cuing, the cues provided at the time of retrieval are analogous to items previously recalled in a free recall situation. Statistically, it becomes more difficult to sample a noncued item as more cues are provided. Combined with the fact that cuing also makes the cued items more salient and therefore more likely to be sampled, cuing may inhibit recall of noncued items.

One additional and critical aspect of this model is called the cessation rule. It states that attempts to recall new items may cease once a criterion number of consecutive samples produces no new items. It follows that the speed with which the criterion is - reached varies directly with the number of cues provided. Thus, as an advertisement describes a larger proportion of a product's attributes, the longer it may take a consumer to recall an unmentioned one and the sooner s/he will stop trying.

The purpose of this research is to determine the degree to which the principle of part-list cuing applies to attribute recall. Attribute information differs from other categorical information in that recall could be guided by the prototype object of which the attributes are a part. For example, instead of sampling memory for automobile attributes in an unstructured, unplanned way, a consumer might image a prototypical car and use it as a very effective recall cue. This strategy could conceivably eliminate the effects of part-list cuing since recall would not involve a simple sampling-with-replacement process. In both experiments reported here, an automobile is used as the stimulus object because consumers are likely to possess prototypes and/or exemplars of it in memory, and because automobiles possess numerous, describable attributes.

EXPERIMENT 1

Three factors were manipulated in the first experiment: number of cues provided at recall, sex, and cue organization. The rationale for the first factor was described above, that is, recall should become more difficult as the number of retrieval cues is increased. In this experiment, the number of cues provided was either 0, 5, or 15. Sex was included as a factor because it was thought that this product class was one with which males and females might be differentially familiar. The part-list cuing effect should be most robust for subjects whose attribute knowledge is relatively unstructured in memory. For subjects who are more knowledgeable, and therefore possess well-organized, categorical memory structures, cues may actually enhance recall if the cues help the consumer access unrecalled categories of information (cf. Tulving and Psotka 1971).

Finally, since part-list cuing seems most effective when the cues and the target recall items are from the same category or subcategory, a third factor was added: across- versus within-category cuing. The stimuli in this experiment described several automobile dimensions (e.g., comfort, performance, safety). The part-list cues presented at recall were sampled from all of the dimensions (across-category) or from only a subset of them (within-category). It was predicted that inhibition would be significant in the across-category condition since no category would be immune from the effect. Less inhibition was expected in the within-category condition since cues from one category should not inhibit recall from other categories, especially when only a small number of categories are present.

Method

Materials. A list of 30 automobile attributes was compiled. The only criteria used were that the attribute descriptions be short enough to fit on slides and that they each be closely associated with a particular higher-order dimension. The target items consisted of 5 attributes relating to each of the following dimensions: safety, comfort, performance, appearance, and economy. The sixth dimension, construction quality, was used to provide buffer items at the beginning and end of the critical list in order to control for primacy and recency effects, and was not counted in the scoring of the data.

At the time of recall each subject was provided with a sheet of paper containing, 3, 5, < 5 list cues. The cues were all of the items contained in 1 or 3 of the categories (within-category condition) or were 1 or 3 items sampled from each of the categories (across-category condition). To control for the differential saliences of the cues, a total of 20 lists were constructed by creating 5 versions for each of the list type (across vs. within) X cue number (5 or 15) conditions. In the 5-within condition, all 5 items from a single category served as the cues. Each of the 5 versions consisted of a different category. In the 5-across condition, one cue was selected from each of the categories. Across the 5 versions, each cue was unique. That is, each category contributed a different item to each list version. In the 15-within condition, 3 categories were randomly sampled from the set of 5, 5 different times. Finally, in the 15-across condition, 3 cues were randomly sampled from each of the categories, 5 different times. In the two latter conditions, the orders in which the categories appeared on the lists were also randomized.

Procedure. Subjects were run in small groups. With the exception of the construction quality dimension, the slides were presented in a blocked fashion, i.e., all items from the same subcategory appeared consecutively. This was done in order to maximize the salience of the categorical nature of the attribute list. The order in which the categories were presented was randomized for each group of subjects. The attributes were projected onto a blank screen at a 3-second rate. Following presentation, subjects were asked to solve a difficult but unrelated problem for one minute. This was done in order to eliminate further the possibility of a recency effect. Afterwards, subjects were presented with 0, 5, or 15 of the stimulus attributes and were asked to recall the remaining ones. Six minutes were allowed for recall.

Subjects. All subjects in this and in the following experiment were volunteer undergraduate students enrolled in marketing courses. In the present experiment a total of 180 subjects participated, with an equal number serving in each of the 3 (cue number) X 2 (sex) X 2 (cue category) cells.

Results and Discussion

Since the provision or list cues reduces the set of remaining items that can be recalled, the dependent variable was the proportion of noncued, non-buffer items recalled by each subject. If the part-list cuing effect occurs, then the proportion of noncued items recalled should vary inversely with the number of cues provided.

The means are presented in Table 1. The data were analyzed using planned contrasts. The results revealed a significant effect of sex (F=4.60, p<.05). Neither the main effect of cue type (F=1.67, p>.10) nor any of the interactions approached significance (all p ' s >.15). As for the variable of primary interest, a significant linear effect of cue number was obtained (F=7.64, p< .01 ), and it was in the predicted direction. That is, recall became more difficult as the number of cues increased.

TABLE 1

The main conclusion to be drawn is that attribute recall can be inhibited through cuing. The failure to find an effect of cue category or a cue X sex interaction could be attributable to a lack of category definition, that is, the cues may not have been perceived as categorical as intended. In an extreme case, a subject would not perceive any category boundaries and would instead process the stimulus information as a see of unrelated items. If so, no effect of cue category would be expected. Although nonsignificant, the results were in the predicted direction. A smaller proportion of items was recalled in the across- than in the within-category condition -- .43 versus .46, respectively. A lack of category definition may have attenuated this effect.

The main effect of sex suggests greater familiarity with the stimuli on the part of males. Since sex did not interact with cue number, however, either the difference in familiarity did not reflect a difference in cognitive structure, or the stimuli used were not precise enough to show one. An analysis of the data showed that males and females recalled a nearly identical number of experimenter-defined categories -- 4.63 vs. 4.57, respectively. Thus, if subjects perceived the category boundaries to the degree intended, they did so equally across sexes.

EXPERIMENT 2

The first experiment demonstrated the general phenomenon of part-list cuing inhibition in terms of overall recall as a function of cue size. In the second experiment, a paradigm that more closely resembles an advertising situation was studied. That is, the effects of cuing on the recall of a single specific attribute was investigated.

Method

Materials. Twenty automobile attributes served as the stimulus set. Of these, four were the object of investigation; they were: (a) 6-cylinder engine, (b) leather dashboard, (c) smooth rite, and (d) low maintenance cost. The four were chosen based on their free recall rate (no-cue condition) in the previous experiment. The first two were among the most frequently recalled; the latter two were recalled infrequently. Note that the frequently recalled items are concrete, physical attributes whereas the other two are not.

Procedure. Subjects were run in small groups. As in the first experiment, the attributes were projected onto a blank screen at a 3-second rate. The order of the attributes was random with the exception that the four critical items appeared in the middle of the list, again in order to control for primacy and recency effects. Also, the attribute presentation was again followed by a one-minute distractor task.

Immediately following the distractor task, the recall test was presented. Subjects in the no-cue (free recall) condition were asked to recall as many of the stimulus attributes as possible. Subjects in the cued condition were presented with a sheet containing 19 of the attributes and were asked to recall the missing one. The missing item was always one of the four described previously. All subjects were given 3 1/2 minutes to complete the task.

Each of the four cue conditions was matched against its own control. For each condition, 12 subjects were cued and 12 were not. Hence, a total of 96 subjects participated.

Results

The number of subjects out of 12 that recalled the critical attribute in each of the four conditions, along with the corresponding X and significance-level values are presented in Table 2. In each case, the critical item was recalled more frequently in the noncued than in the cued condition. In three out of four cases, the difference was significant.

TABLE 2

In the one case that failed to reach significance, "smooth ride," it is apparent that there is a floor effect; only 3 subjects in the control condition were able to recall the item. Consequently, it is not meaningful to make an overall comparison of the concrete and abstract attribute conditions. Nonetheless, it is safe to conclude that cuing at the time of recall or attitude formation may inhibit recall of a specific attribute, and that the effect is not limited to physical attributes.

GENERAL DISCUSSION

Previous studies have demonstrated that a problem may be framed by the manner in which attribute information is presented in advertisements. In those studies, problem framing was accomplished by focusing consumers' attention on specific attributes at the time of initial learning. This was made possible by using subjects that were novices with respect to the product class (Wright and Rip 1980) or by using attributes with which the decision makers were unfamiliar (Gardner 1983). The present research has shown that a problem also may be framed at the time of recall, independently of attentional factors, and with products and attributes that are highly familiar. Further, these studies a]so demonstrate that low involvement is not necessary condition for problem framing to occur through advertising (cf. Krugman 1965). Even under high involvement, consideration of particular attributes may be inhibited. The result is an incomplete evaluation of a product and a relative increase in the importance assigned to attributes that are retrieved -- effects that previous studies have demonstrated by manipulating learning and attention.

This is not to say that involvement plays no role in the part-list cuing paradigm. It was noted previously that according to the cessation rule of the Rundus Model, sampling from memory may cease once a criterion number of samples produces no new information. It may be the case that the value at which the criterion is set is influenced by the consumer's involvement in the decision. That is, the criterion value may vary directly with involvement level. Under low involvement, consumers may cease trying to recall new information after a relatively few unsuccessful recall attempts. A highly involved consumer may remain undeterred until an exhaustive search has been attempted.

It was also noted at the outset that part-list cuing is a case of a more general recall phenomenon in which retrieval becomes more difficult as more items are recalled. This general phenomenon is known as output interference (Roediger 1974). In most cases of attribute recall, this phenomenon should not pose a serious problem since consumers should first recall the attributes they consider to be most important. Since the attributes that would experience the greatest amount of interference would be recalled last, the decision process may not be adversely affected to any great extent. Nonetheless, the decision may not be a totally informed one. Consider, for example, the use of decision plan nets (Park 1982). If a consumer is asked to free recall all dimensions of importance in their order of importance, those that are least important may not be retrieved. At the margin, this could affect predictions about the decision the consumer will make. Similarly, if the consumer is provided with a list of important dimensions and told to add any others s/he would consider, then the typical part-list cuing paradigm obtains. Therefore, regardless of methodology, some dimensions may not be considered due to recall inhibition.

It should be noted that the part-list cuing or output interference effects may also apply to attribute alternatives as well as attribute dimensions. That is, if a particular dimension such as style has many possible alternatives, recall of a particular alternative may be inhibited through the mechanisms discussed above.

Finally, Gardner (1983) has correctly argued that the use of problem-framing strategies by advertisers has policy implications. Regardless of the locus of the effect (i.e., at encoding or retrieval), problem framing may result in suboptimal decision-making. It would seem that intentional use of such strategies to distort the decision process could be remedied by a policy of full disclosure.

REFERENCES

Boyd, Harper W., Jr., Michael L. Ray and Edward C. Strong (1972), "An Attitudinal Framework for Advertising Strategy," Journal of Marketing, 36 (April), 27-33.

Gardner, Meryl Paula (1983), "Advertising Effects on Attributes Recalled and Criteria Used for Brand Evaluations." Journal of Consumer Research, 10 (December). 310-318.

Krugman, Herbert E. (1965), "The Impact of Television Advertising: Learning Without Involvement," Public Opinion Quarterly, 39 (Fall), 349-356.

Lynch, John G., Jr. and Thomas K. Srull (1982), "Memory and Attentional Factors in Consumer Choice: Concepts and Research Methods," Journal of Consumer Research, 9 (June), 18-37.

Park, C. Whan (1982), "Joint Decision in Home Purchasing: A Muddling-Through Process," Journal of Consumer Research, 9 (September), 151-162.

Roediger, Henry L., III (1974), "Inhibiting Effects of Recall," Memory and Cognition, 2 (April), 261-269.

Rundus, Dewey (1973), "Negative Effects of Using List Items as Recall Cues," Journal of Verbal Learning and Verbal Behavior, 1: (February), 43-50.

Tulving, Endel and Zena Pearlstone (1966), "Availability versus Accessibility of Information in Memory for Words," Journal of Verbal Learning and Verbal Behavior, 5 (August), 381-391.

Tulving, Endel and Joseph Psotka (1971), "Retroactive Inhibition in Free Recall: Inaccessibility of Information Available in Memory Store," Journal of Experimental Psychology, 87 (January), 1-8.

Wright, Peter and Peter Rip (1980), "Product Class Advertising Effects on First Time Buyers' Decision Strategies," Journal of Consumer Research, 7 (September), 61-64.

----------------------------------------