Chronometric Analysis: an Introduction and an Application to Low Involvement Perception of Advertisements

Meryl P. Gardner, (student), Carnegie-Mellon University
Andrew A. Mitchell, Carnegie-Mellon University
J. Edward Russo, University of Chicago
ABSTRACT - The methodology involved in using response times to examine internal information processing is briefly reviewed and the application of this approach to consumer research is discussed. A study is described that uses response times to examine the differential effects of high and low involvement learning of advertising messages. Low involvement learning is hypothesized to take place under either attention limited or strategy limited conditions. The experiment examined the latter condition where consumers fully attend to the advertisement but not for the purpose of evaluating the advertised good. The results of the experiment confirm several predicted effects of this type of low involvement processing, including the formation of more positive attitudes toward the advertised brand.
[ to cite ]:
Meryl P. Gardner, Andrew A. Mitchell, and J. Edward Russo (1978) ,"Chronometric Analysis: an Introduction and an Application to Low Involvement Perception of Advertisements", in NA - Advances in Consumer Research Volume 05, eds. Kent Hunt, Ann Abor, MI : Association for Consumer Research, Pages: 581-589.

Advances in Consumer Research Volume 5, 1978      Pages 581-589


Meryl P. Gardner (student), Carnegie-Mellon University

Andrew A. Mitchell, Carnegie-Mellon University

J. Edward Russo, University of Chicago


The methodology involved in using response times to examine internal information processing is briefly reviewed and the application of this approach to consumer research is discussed. A study is described that uses response times to examine the differential effects of high and low involvement learning of advertising messages. Low involvement learning is hypothesized to take place under either attention limited or strategy limited conditions. The experiment examined the latter condition where consumers fully attend to the advertisement but not for the purpose of evaluating the advertised good. The results of the experiment confirm several predicted effects of this type of low involvement processing, including the formation of more positive attitudes toward the advertised brand.


All consumer decisions rely, to some extent, on information stored in memory. Recent research on consumer decision making, however, has de-emphasized the role of memory in consumer decision processes by concentrating attention on external information acquisition. Since acquisition responses, such as reaching or looking, are conveniently observable, they offer a clear methodological advantage. In contrast, retrieval of information from memory is covert.

It seems clear, however, that if we are to understand consumer decision processes the important role of memory and internal processing cannot be excluded from experimental investigation. Currently, a consensus seems to be forming that the time is appropriate for consumer research to direct its attention at the internal processes of information encoding and storage and to their effect on consumer decision making (Bettman, 1978; Chestnut and Jacoby, 1977; Johnson and Russo, 1977; Olson, 1978).

Available Methodologies

The investigation of internal processes is hindered by the lack of associated overt behavior. There are, however, two observations that can always be made on a cognitive process: the output of that process and the time taken to complete it. Both of these measures provide a data base for methodologies that can successfully investigate internal processes. The focal point of this paper is on chronometric analysis or the use of response times to examine internal processes; however, it might be helpful to first discuss methodologies based on output. To use output effectively, the experimental environment has to be designed so that the observed response can discriminate among possible processes. This is called input-output analysis, or, in its most refined form, the axiomatic method (Chase, 1977).

The output of a process can be elaborated by two process tracing techniques. The more common requires subjects to generate a verbal report of their own cognitive processes. Technically, these verbal protocols are the output of a second process, that of self-commentary. Although often very informative, they are difficult to analyze formally and are susceptible to distortion when based on memory rather than emitted concurrently with the primary task. Furthermore, verbal protocols are almost useless for brief or highly automatized cognitive processes of which subjects are not consciously aware (Nesbitt and Wilson, 1977). Unfortunately, much consumer behavior, such as encoding information from advertisements, may be of this nature.

A second elaboration of simple output analysis is the recording of intermediate responses, such as information acquisition behavior. This is still output analysis, except now the output of subprocesses is analyzed alongside that of the primary process. Recent consumer research has made excellent use of this technique. Through its application we have learned that people often ignore large amounts of product information, they adapt to the presentation mode or format of the information and they are flexible in their selection of decision strategies, seldom using a single, "pure" strategy. This progress has been achieved, however, in laboratory settings characterized by external sources of product information, and, when investigating known alternatives, has failed to take into account previously stored information (Bettman, 1978).

Apart from these two process tracing elaborations, the input-output framework includes a wide variety of specific experimental paradigms for the investigation of memory. Tulving and Bower (1974) provide a list of ten such techniques commonly used in cognitive psychology. These include: feature probing, recall intrusions, retrieval cueing, and clustering in recall. Although we shall now turn out attention to chronometric analysis, the power and diversity of input-output analyses should not be ignored.


As mentioned previously the purpose of chronometric analysis is to use response times to examine internal information processes. In this context, response times (RT's) are generally defined as the amount of time between the presentation of a stimulus to an individual and his or her response to that stimulus.

Chronometric analysis has been used extensively by cognitive psychologists to examine a wide range of mental processes. It has been used, for instance, to examine the underlying mental processes involved in verifying that a presented stimulus belongs to a previously memorized set (Chase and Clark, 1972; Clark and Chase, 1972) and the quantification of visual stimuli (Klahr and Williams, 1976).

The fundamental principle of chronometric analysis is that response time measures the amount of processing that has taken place. In order to do this we generally have to assume that the processes are executed in series rather than in parallel. For many applications this principle alone is sufficient. Subjects, for instance, may be asked to respond to the same stimulus under different conditions that imply different response times. The observed RT's are then used to test the theory from which the predictions were derived.

Before we discuss general procedures for using chronometric analysis to examine internal processes, it will be helpful to present a specific example of its use from the cognitive psychological literature. Collins and Quillian (1969) used response times to test Quillian's (1969) model of semantic memory. According to the model, semantic memory can be represented as a hierarchical network with each node representing an object or concept. An example of such a hypothetical semantic network [In Collins and Quillian's experiment, the structure of semantic memory for animals was examined by using the subsets of birds and fish.] for cola and lemon lime soft drinks is presented in Figure 1. Moving to a higher level in the network signals a superset relation. For example, cola is the superset of Pepsi Cola, and soft drink is the superset of cola.



The power of this conceptualization is both the efficiency of information storage, since information is stored at its highest level of generality and the ability to infer or verify novel facts. For example, an individual does not have to store both "Coca Cola has a dark color" and "Pepsi Cola has a dark color"; this information is stored with the concept cola. Also, the sentence "Pepsi Cola has a dark color" can be verified even though that particular association has never been encountered.

According to the model, the verification process requires a search through the nodes of the semantic network. For instance, verifying "Coca Cola has a stimulating taste" would require activating the links from Coca Cola to cola to soft drink where the property "has a stimulating taste" can be accessed.

To test the model, RT's were recorded as subjects verified statements that involved searching different levels of the hierarchy. In general, the results supported the model. [The data for positive responses fit the model well, however, this was not true for negative responses. It was hypothesized that different strategies may be involved in falsifying sentences.] The more hierarchies that needed to be searched to respond to a question, the longer the response time. (Other models of memory have since been shown to be compatible with the results; see Chase, 1977 [For instance, distance in semantic network is confounded with past associative frequency and, therefore, with associative strength.]). The response times also provided estimates of the time needed to traverse between nodes (about 75 msec) and to retrieve a property at a node (about 225 msec). The former can be measured by subtracting the times to verify sentences like "Sprite is lemon lime" from "Sprite is a soft drink" or "Sprite is made by Coca Cola" from "Sprite has a clear color." Each pair of sentences involves the same processing except for an additional level change in the second sentence.

The Subtractive Method

The Collins and Quillian experiment illustrates the use of the subtractive method which was first proposed by Donders (1868). Logically the method is quite simple. If a number of mental processes occur serially, then by eliminating one or more of the processes in a particular experimental condition, one may obtain a measure of the time required for the remaining processes. In the Collins and Quillian experiment, the serial operations are an original retrieval of a concept from memory and then a sequential search through the different hierarchies. The occurrence of this relationship in an actual experiment provides evidence supporting the hypothesized model.

The validity of the subtractive method relies on an assumption that might be called the principle of process Deletion. It is assumed that the process of interest can be deleted from some sequence of processes without affecting any of the others. That is, the deleted process does not interact with any other process. (Technically, the durations among processes must not interact.) This assumption is critical to the validity of the subtractive method but, unfortunately, is not always testable.

From the above discussion, it should be clear that the subtractive method requires an adequate theory of the process that is to be examined. In order to construct a task that deletes only the specific process of interest, one should know the sequence of cognitive activities that occurs between stimulus and response. At the least, one must know enough to be certain that the Principle of Process Deletion is applicable, but usually this knowledge is only part of a theory of the complete process.

The Additive Factor Method

The additive factor method, proposed by Sternberg (1969), is an elaboration of the subtractive method. Its principal concern is to provide procedures for developing a conceptual model of the cognitive processes that intervene between stimulus presentation and the subjects' response. Whereas the subtraction method requires a well-developed theory in order to interpret the resulting reaction times, the additive factor method is designed to aid in the development of a theory of the intervening processes.

The basic logic underlying this method is that each reaction time interval contains a series of independent stages. Each stage of the process transforms the information received from the previous stage and outputs it to the next stage. It is assumed that the duration of each stage is independent of the stages that precede it. The implications of this conceptualization are as follows. First, the total reaction time of a particular process is the sum of the stage durations. Second, independent effects on total reaction time will be produced whenever experimental manipulations (independent variables) affect different stages. Finally, if the effects of two experimental manipulations interact in a statistical sense then they must affect the same stage. Thus, in using the additive factor method, a multifactor experiment is applied to a particular information processing task. Each pair of factors that have an additive effect on reaction time affect different stages of the process, and each pair that interact affect the same stage. Consequently, in order to obtain a precise analysis of the stages for definitional purposes, more than a few factors must be manipulated.

To illustrate the additive factor method, we rely on the experiments of Sternberg (1969) which investigated short term character recognition. In these experiments a subject was required to memorize a set of stimuli and then, upon presentation of a stimulus, to respond positively or negatively depending on whether it was contained in the previously memorized set. The factors manipulated in these experiments included: pictorial clarity of the presented stimulus, size of the previously memorized set, whether a positive or negative response was required, and the relative frequency of either a positive or negative response. The results of these experiments indicated that all four factors had a significant effect on reaction time and that all six pairwise interactions were either insignificant or there was a compelling priori reason for not expecting a statistically significant interaction. Consequently, Sternberg (1969) hypothesized that the task involved four independent stages, each associated with one of the factors tested: encoding, comparison, response choice and response execution. Note that four stages are the most that could be discovered in a four factor design and that there could possibly be more stages to the process that could not be identified in this experiment.

Methodological Problems

As with any technique, chronometric analysis has accompanying methodological problems. The primary problem ia the natural trade-off between speed and accuracy. Subjects can usually perform faster if they sacrifice accuracy or more accurately if they sacrifice speed. This relation is captured by the speed-accuracy operating characteristic (Pew, 1969). An idealized version of this function, adapted from Pachella (1974), is shown in Figure 2.



According to the definition of reaction time, subjects should be responding at point A. However, they generally operate below this point. Most of the problems associated with the speed-accuracy trade-off can be avoided if subjects perform at the same accuracy level under all experimental conditions. Then the relative values of the observed RT's will reflect the effect of the independent variables and not the effect of a different accuracy strategy. When accuracy fails to remain constant across conditions, there is a danger that incorrect inferences will be drawn from the RT data. (For an example, see Pachella, 1974). A number of procedures have been proposed for adjusting reaction times if different accuracy levels occur in an experiment. A critique of these procedures may be found in Pachella (1974).

The second problem to be discussed concerns the interpretation of response times. Whenever detailed processes are manifested only in behavior as aggregate as response time, the theoretical conclusions must rely heavily on the theoretical input. In spite of the power of the additive factors method, it still requires reasonable hypotheses about potential stages of the process in order to design an appropriate multifactor experiment.

Prior to Sternberg's introduction of the additive factors method, the subtractive method had fallen in disfavor. In discussing this particular period, Chase (1977) states, "In retrospect, it appears that the lack of success with the subtractive technique was due primarily to the absence of good cognitive theory." For consumer researchers interested in relatively complex tasks, this means that RT's are not the proverbial free lunch. Even the most sophisticated chronometric analysis can only refine theory, not create it.

Response Time and Consumer Behavior

There are a number of different areas in consumer behavior where the use of reaction times would seem to be useful. One area is the structure of product knowledge (i.e., the organization of memory for product information). A number of questions are of interest here. What is the structure linking brands and attributes within a product class? Is it hierarchical as pictured in Figure 1, or organized in accord with the more recent models of semantic memory (e.g., Anderson and Bower, 1973, or Norman and Rumelhart, 1975)? Does brand name play a central role linking all brand associations, as some researchers have argued? What is the relative efficacy of brand-based versus attribute-based memory organizations? When the structure of product knowledge is identified, will it be invariant across individuals and product categories? These are only some of the more obvious questions that need to be answered about product knowledge, a major understudied area of consumer research.

The use of chronometric analysis to explore the structure of product knowledge has recently been demonstrated by Johnson and Russo (1977). The duration of a cued retrieval task identified the distance in the memory structure between the cue and the target item. This method was used to show that memory structure was largely determined by the input organization of the product information. Interestingly, some reorganization of the structure did occur, and it appeared to favor storage by attribute.

Just as important as the structure of product knowledge is the content of that knowledge. In a study of multiple decision making on a single information set, Green, Mitchell and Staelin (1977) found that subjects do not examine all the available information and more importantly, they are able to retrieve only a small portion of the examined information in memory. Examining what information is stored, in what form and how it enters memory are natural problems for response time techniques.

The importance of the content issue can be demonstrated by noting that either verbal report techniques or self administered scales have traditionally been used in most applied and theoretical research for identifying the contents of semantic memory. Of interest is how much of this reported information is based on stored information and how much on inference. While the retrieval of stored information is apt to be similar in the laboratory and in a real purchase situation, the inference processes may be very different in these two contexts.

A third use of reaction times is for examining the development of cognitive structures. Familiarity with the product class and experience with a brand have been shown to be important determinants of both purchase behavior and evaluation processes. For example, Dover and Olson (1977) have presented evidence that the predictability of the Fishbein model changes as experience with a brand changes. Psychological theory suggests that learning about a particular concept both strengthens and adds components to a cognitive structure until it becomes unitized, i.e., may be considered a single chunk of information (Hayes-Roth, 1977). When this occurs, retrieval of information from the cognitive structure will be much quicker than at earlier stages.

In a pretest for one of our experiments (Edell and Mitchell, 1977) one group of subjects was exposed to product information. A second group received this information and, in addition, was allowed to use sample products. The resulting cognitive structure of the second group enabled the retrieval of the identical information to be completed, on the average, a full second faster than for the first group.

The previous illustrations have suggested the use of response times to measure the structure, content, or development of cognitive structures for storing product knowledge. Many of the questions raised can be answered with the simplest use of total RT's, without partitioning the process into stages or subprocesses. Others, such as determining the subprocesses involved in information acquisition, will require the use of the subtractive or additive factors methods. Although there is considerable opportunity for fruitful application of these various RT techniques to consumer research, we conclude this section with a few caveats to overenthusiasm.

Many areas of consumer behavior do not yet have process-oriented theories with the strength to support chronometric experimentation. If one cannot start without a reasonable idea of the processes being examined, no methodology, chronometric analysis included, is capable of single-handedly generating truth. Second, using response times to identify the information retrieved from memory versus that based on inferences may prove very difficult. This is especially true if retrieval times exhibit large variances both between and within individuals, as they usually do. Third, unlike artificial laboratory environments, many questions appropriate to consumer situations do not have correct answers. This will cause difficulties in adjusting RT's for the effects of a speed-accuracy trade-off. Finally, note that chronometric analysis is a sophisticated methodology developed by experimental psychologists over a period of many years. Strong behavioral and experimental expertise is required of researchers who want to benefit fully from the application of this technology to consumer behavior.


Low involvement learning of advertising information was first hypothesized by Krugman (1965) over a decade ago. Since that time the concept has generated considerable speculation (e.g. Ray, 1973), but almost no empirical research. Perhaps one reason for the dearth of empirical research is that we do not have a precise definition of low involvement learning in terms of cognitive processes. One of the goals of this study is to provide such a definition.

High Involvement Learning

We might begin the task of defining low involvement by first considering what occurs in high involvement learning. Here we hypothesize that interest in the product category is high and the consumer is actively processing the information in the advertisement to reach an overall evaluation of the advertised brand. For example, individuals intending to purchase stereo components may actively seek and read advertisements to evaluate alternative components or vendors. In this situation the consumer has a goal, evaluation of alternative components, which determines the processing strategy that is used during exposure to an advertisement. Under these conditions, we would expect the consumer to critically evaluate the information about the brand from the advertisement by generating thoughts or cognitive responses (Wright, 1973, Greenwald, 1968).

Low Involvement Learning

Alternatively, no evaluative processing may take place during exposure to an advertisement. This may occur, for instance, when an individual becomes involved in the photograph of a print advertisement for a brand in a product category that he or she does not intend purchasing. Alternatively, this may also occur when an external stimulus diverts attention from the processing of information contained in the advertisement. Since in both of these situations little evaluative processing is taking place, we would not consider them examples of high involvement learning.

Based on these examples, let us distinguish between two causes of low involvement processing of advertising messages: attention limitations and strategy limitations. Processing is attention limited when the advertisement does not receive enough attention for it to be fully perceived or evaluated. Following Kahneman (1973) or Norman and Bobrow (1976), attention is considered to be a limited cognitive resource. In other words, the system for processing information at any given point in time has limited capacity. Consequently, insufficient attention can be caused by a deliberate diversion of attention to other processes. For instance, an ongoing conversation with another person may divert whatever attention the individual was devoting to a radio advertisement. This type of low involvement learning has been termed distraction and has been studied by psychologists (Festinger and Maccoby, 1964) and within a marketing context (Bither, 1972). Theoretically, the diversion of attention to other processes limits the attention devoted to the processing of information from the advertisement and, consequently, reduces the number of thoughts or cognitive responses that might be generated. The usual research result under this paradigm is that more positive attitudes are formed with distraction than without. Notice that the attention limitation may be caused by either the individual or the environment.

Low involvement learning can also occur in situations of full attention, if that attention is not devoted to an evaluative processing of the advertising message. Recall the earlier example of attention devoted to the photograph in a print advertisement rather than to brand related information. We call this strategy-limited low involvement because the person processes the advertisement with other than an evaluative strategy. Under these conditions few, if any, cognitive responses about brand information from the advertisement would be generated during exposure to the advertisement with little or no resulting effect on the contents of semantic memory for the brand. However, since the individual was devoting full attention to the advertisement, a trace of the advertisement would be stored in episodic memory (Tulving, 1972). At some later point, information from the advertisement might be retrieved from episodic memory and transferred to semantic memory. However, during this transfer process, we hypothesize that the information would not be as critically evaluated as it would be if it had been received directly from the environment under conditions of high involvement. Consequently, we would expect differences in the resulting cognitive structures after exposure to the same advertisement under the two conditions. More specifically, we would expect more positive attitudes to be formed under conditions of strategy limited processing as opposed to high involvement processing.

This phenomena of strategy limited low involvement may be especially prevalent in the broadcast media where effort is required to avoid exposure to an advertisement. In situations where the viewer has little interest in the brand being advertised, he or she may resort to nonevaluative processing instead of exposure avoidance.

Several further points should be noted. First, both attention-limited and strategy-limited low involvement can occur simultaneously. The same advertisement can experience low involvement for both reasons. Second, in neither low involvement situation is it claimed that no evaluative processing occurs. Instead, partial evaluative processing can, and probably does, occur under both attention-limited or strategy limited conditions. This partial processing, combined with repeated exposure, can lead to the learning of the advertising message. Third, the low and high involvement conditions described previously should be viewed as end points on a continuum. One example of a point along this continuum might be the case where an individual processes information from an advertisement to learn about an alternative, but he or she does not choose to make an overall evaluation of the brand at this point. Fourth, the analysis of low involvement learning presented above is constructed within an information processing framework. The emphasis is on the cognitive processes that are active during the perception of an advertisement. The subject state or even the intentions of the individual are relevant only in as much as they affect those processes.

Since attention-limited low involvement learning has been studied previously, we direct our efforts toward strategy limited low involvement learning under the condition of full attention.


The purpose of this study is threefold. First, we wanted to demonstrate the nature of processing under strategy limited low involvement conditions and contrast it to processing under high involvement conditions. Second, we wanted to examine the differential effects on the re-suiting cognitive structures of individuals exposed to the same advertisements under these two conditions. Finally, we wanted to demonstrate the usefulness of chronometric analysis in examining the processing that takes place during exposure to advertisements.

Consequently, we wanted to design an experiment where the manipulation would be the creation of the two different involvement conditions during exposure to a number of advertisements. The difference in processes under the two conditions would be demonstrated by chronometric analysis. Subjects would be asked to verify statements which would require different mental processes and, therefore, different response times, under the two conditions. Finally, measures would be taken of the resulting cognitive structures of the brands featured in the advertisements.

The experiment should be viewed as an exploratory study whose purpose was to provide insights into the effects of low involvement processing of advertising information and the general cognitive processes involved in the learning of information from advertising.



High involvement was produced in the experiment by instructing subjects to examine the advertisements as though they were planning a purchase of the product class of the brand in the advertisement. To enforce this evaluation task, subjects were required to think aloud while making their evaluations and if they did not reach an overall evaluation of the brand, they were prompted to do so.

The low involvement task involved evaluating the advertisements as to their ability to attract and hold attention. Since subjects would generally be unfamiliar with this task, a number of criteria were given to them which they were to use in making their evaluations. These criteria were the amount of onomatopoeia, assonance, alliteration, rhyme, hyperbole in the copy and the number of times the words "you" and "your" appeared. Subjects were also required to think aloud while making their evaluations and again, if they did not reach an overall evaluation of the advertisement, they were prompted to do so.

Statements were developed whose verification should require different cognitive mental processes under each experimental manipulation. Since little is known about these processes our model was necessarily simple. The high involvement group was expected to form both brand and attribute evaluations during exposure to the advertisements, therefore, the only cognitive processes required in verifying these statements should be the direct retrieval of information from semantic memory. For the low involvement group, verification of similar statements should require both retrieval of the information contained in the advertisement from episodic memory and then additional processing of this information to form attribute and brand evaluations. Therefore, the response times to these statements should be longer for the low involvement group than the high involvement group.

Similarly, the low involvement group should form evaluations with respect to the different "attention" criteria and an overall evaluation of the advertisements' ability to attract and hold attention. Verification of these types of statements should require only the direct retrieval of this information from semantic memory. The high involvement group, however, will need to retrieve the trace of the advertisement from episodic memory and then evaluate the advertisement according to the stated criteria in order to verify these statements. Therefore, the response times for the high involvement group should be longer than those for the low involvement group on these statements. An example of each type of statement is given in Table 1.




An after only design was used with the main manipulation being the two different levels of involvement. Two different dependent variables were examined. The first was the response time to the four different types of statements: brand evaluations, attribute evaluations, attention evaluations and attention criteria evaluations. Advertisements for four different brands from four different product categories were used. For each brand, one brand evaluation, one attention evaluation, two attribute evaluation and two attention criteria evaluation statements were used.

The second dependent variable was the attitude toward each brand. The mean score of three bipolar evaluative scales (good-bad, dislike very much-like very much, poor quality-high quality) was used for this measure. These evaluations were obtained from a questionnaire administered after the response time statements.


The product categories selected for the advertisements were automobiles, pens, tires and shoes since these types of products would be familiar to our subjects. The copy for each advertisement was designed to provided factual information about the brand so that an overall evaluation of the brand along two different attributes, such as "riding comfort" for the automobile, could be made by subjects. The copy was then given to an artist who made an ink sketch of a picture to go with the advertisement. Each advertisement contained a picture, headline below the picture and then copy below the headline. The advertisements were not of professional quality. It was obvious to the subjects that they were constructed by the experimenters.


The subjects were 30 individuals of both sexes recruited from the campus community. The sample included students, secretaries and staff members. The payment of subjects was based on the number of correct responses to a subset of response time statements. This subset includes attribute evaluations and attention criteria evaluations. Subjects received 10~ for each correct response and lost 10~ for each incorrect response. Any subject who earned less than $1.00, based on the number of correct and incorrect responses, was given m dollar.


Response times were measured on a tachistoscope. The questions were placed at one end of the apparatus. Subjects looked into the tachistoscope at the other end, but could not see the questions until they pressed the starter pedal. This simultaneously exposed the question and started the timer. The subject answered the question by pressing either a "true" or "false" button which stopped the timer.


Subjects were alternately assigned to low and high involvement conditions to achieve randomization and were run one at a time. At the beginning of the experiment subjects were given instructions depending on their assigned experimental condition and then given two advertisements to evaluate so they would become familiar with their respective tasks. Next they were shown the four advertisements one at a time and asked to think aloud as they did their evaluations. These verbal responses were recorded on a tape recorder.

Subjects were then directed to the tachistoscope and given instructions as to its use. They were then given ten sample statements, completely removed from the context of the experiment, to allow them to become familiar with the operation of the machine. Prior to the experimental questions, subjects were informed that they may receive some questions that they might not be prepared to answer but that they could do so if they thought back to the advertisements. They were also told that factual statements in the advertisement were true, that they should attempt to answer all questions with equal accuracy, and were informed of the payment schedule. Finally, subjects in the high involvement task were given the criteria for evaluating each advertisement on its ability to attract and hold attention.

The statements were asked in the same order for all subjects. The order was: brand evaluation statements, attention evaluation statements, attribute evaluation statements, and attention criteria statements. Within each statement type, the order of the products was also the same: car, pen, tire and shoe.

After answering the questions on the tachistoscope, the subjects were given a questionnaire to complete and after this they were debriefed and paid. The entire procedure took approximately an hour for each subject.

Data Analysis

Response time data are almost always skewed upward (Pachella, 1974). In this experiment, the average response time over all subjects and conditions was 4.45 seconds, yet some of the observations were as large as 23 seconds. In these instances, processes other than direct retrieval or inference were being executed.

In analyzing response times with these characteristics, two approaches are commonly employed (Pachella, 1974). Either medians are substituted for means or all observations above some cutoff value are dropped. This initial analysis of the data took the second approach. All observations that were twice as long as the average were eliminated. The resulting means were also compared to the median of all the observations for a particular cell.

When two questions were used for a particular question type, these two observations were averaged to produce a single observation for a subject. If one of the response times was greater than the cutoff value, then the response time for the other question was used for that observation.


Response Times

The main prediction of the experiment is the faster verification of brand evaluation statements by the high involvement group. Similarly, the low involvement group should exhibit faster verification of attention evaluation statements. The mean response times for these two types of statements confirm both predictions. These data are plotted in Figure 3. A one tailed test of the means of the brand evaluation statements (3.54 vs. 4.61) was significant (p < 0.001) as was the same test on the attention evaluations (4.32 vs. 3.80; p < 0.05).

The mean response times for attribute evaluation and attention criteria evaluations are plotted in Figure 4. As predicted, the high involvement group verified attribute evaluation statements faster (3.53 vs. 4.00) and the low involvement group verified attention criteria statements faster (5.46 vs. 4.61). A one tailed test on the means of these two statement types was significant for both the attribute evaluation statements (p < 0.05) and attention criteria statements (p < 0.005). Notice also that, in general, it took longer for both groups to verify attention criteria statements than attribute evaluation statements.






The median scores for each group by statement type is given in Table 2. All of the predicted effects, except for attention evaluations, are confirmed. An examination of the data indicated that two subjects in the low involvement group were largely responsible for this result. Their response times to attention evaluation statements were consistently above 8 seconds which caused the median of this group to be almost equal to the median response of the high involvement group on this type of statement.



In general, the mean response times, after eliminating observations greater than 9 seconds, are larger than the medians. This indicates that in these cases the data is still skewed.

Correct Responses

As mentioned previously, the attribute evaluation and the attention criteria evaluation statements were structured so there was a correct response. We hypothesize that the high involvement group should be more accurate on attribute evaluation statements and on attention criteria evaluation statements, the low involvement group should be more accurate. This prediction held for the attribute evaluation statements, where the high involvement group verified a 77% of the statements correctly, while the low involvement group verified 63% correctly. This difference is significant for a one tailed test (p < 0.05). The high involvement group performed slightly better than the low involvement group on the attention criteria statements (55% vs. 52%), however, this difference is not significant (p < .25).

Attitude Toward the Brand

A 2x4 ANOVA on each subject's overall attitude toward the brand (Table 3) resulted in both a main effect due to group (p < 0.01) and product (p < 0.001). These main effects indicate a systematic difference in attitudes between groups and products. The group-brand interaction was not significant. A plot of the mean attitude scores for each brand and group is presented in Figure 5. It should be noted that the attitude toward each brand is more positive in all cases for the low involvement group.




In this paper we have discussed the use of chronometric analysis to examine internal information processing. Different approaches for using the technique were presented, along with methodological problems and finally suggestions of areas in consumer behavior where the technique might be applied.



An experiment examining the effects of high and low involvement learning of advertising information which utilized response time measures was described. Two different types of low involvement learning were hypothesized. The first, attention limited, occurs when only partial attention is directed at advertisements and has been examined previously as distraction effects. The second, strategy limited, occurs when individuals devote full attention to an advertisement, but do not process the information contained in the advertisement to reach overall evaluations.

In this study, strategy limited low involvement learning under the condition of full attention was examined. Response times were used to examine the retrieval of different types of information from memory under the two conditions. The pattern of the response times between groups for different statement types indicated that the high and low involvement groups processed information from the advertisements differently. Furthermore, the data indicated that more positive attitudes toward the brands were formed under low involvement conditions.


The preliminary results from the experiment illustrate many of the problems in using chronometric analysis in general and to examine consumer behavior specifically. First, the distribution of response times for each statement type within groups was highly skewed. Generally, observations that are considerably greater than average indicate that the subject, in that particular instance, was executing a very different process. These occurrences must be taken into account in the data analysis. In our initial analysis, response times greater than twice the overall average response time were eliminated. Second, there was some indication that the two groups may have been operating on different parts of the speed-accuracy operating characteristic. The low involvement group had slightly lower accuracy on the attention criteria statement than the high involvement group although it had been hypothesized that they should have had greater accuracy. This may be an indication that the response times for this group had a downward bias. Finally, the results indicated that a model of the process that is more sophisticated than the one initially postulated may be required. For instance, more time was required to retrieve attention criteria information than attention criteria information. Our simple model did not provide an explanation of these differences.

In summary, the experiment indicated that low involvement learning of advertising can occur and that in our experiment more positive attitudes were formed with low involvement learning than with high involvement learning. We are now in the process of doing additional data analysis to understand why more positive attitudes were formed under low involvement conditions and also examining the amount of learning of information under each condition.


J. R. Anderson and H. Bower, Human Associative Memory, (New York: John Wiley and Sons, 1973).

J. R. Bettman, An Information Processing Theory of Consumer Choice, (Reading, Mass.: Addison-Wesley, 1978), forthcoming.

S. W. Bither, "Effects of Distraction and Commitment on the Persuasiveness of Television Advertising," Journal of Marketing Research, 9 (1972), 1-5.

W. G. Chase, "Elementary Information Processes," in William K. Estes (ed.), Handbook of Learning and Cognitive Processes Vol. 5, (Hillsdale, NJ: Lawrence Erlbaum, 1978), in press.

W. G. Chase and H. H. Clark, "Mental Operations in the Comparison of Sentences and pictures," in L. Gregg (ed.) Cognition in Learning and Memory, (New York: John Wiley and Sons, 1972).

R. W. Chestnut and J. Jacoby, "Consumer Information Processing: Emerging Theory and Findings," in Arch G. Woodside, et. al. (eds.), Consumer and Industrial Buying Behavior, (New York: Elsevier North-Holland, Inc., 1977.

H. H. Clark and W. G. Chase, "On the Process of Comparing Sentences Against pictures," Cognitive Psychology, 3 (1972), 472-517.

A. M. Collins and M. R. Quillian, "Retrieval Time From Semantic Memory," Journal of Verbal Learning and Verbal Behavior, 8 (1969), 240-248.

F. C. Donders, "On the Speed of Mental Processes," Acta Psychologia, 30 (1969), 412-431. Translated from the original by W. G. Koster from Onderzoekingen gedaan in het Psysiologisch Laboratorium der Utrechtsche Hoogeschool, 1868, Tweede reeks, II, 92-120.

B. A. Dosher, "Retrieval of Sentences from Memory: A Speed-Accuracy Study," Cognitive Psychology, 8 (1976), 291-310.

P. A. Dover and C. Olson, "Dynamic Changes in the Expectancy-Value Attitude Model as a Function of Multiple Exposures to Product Information," in B. A. Greenberg and D. A. Ballenger (eds.), Contemporary Marketing Thought, (Chicago: American Marketing Association, 1977), 455-460.

J. Edell and A. Mitchell, "The Mediating Effect of Product Class Familiarity Upon Cognitive Responses to Advertising,'' Working Paper, Graduate School of Industrial Administration, Carnegie-Mellon University, 1977.

L. Festinger and N. Maccoby, "On Resistance to Persuasive Communications," Journal of Abnormal and Social Psychology, 68 (1964), 359-366.

R. Green, A. Mitchell and R. Staelin, "Longitudinal Decision Studies Using a Process Approach: Some Results From a Preliminary Experiment," in B. A. Greenberg and D. A. Ballenger (eds.), Contemporary Marketing Thought, Chicago: American Marketing Association, 1977), 461-466.

A. G. Greenwald, "Cognitive Learning, Cognitive Response to Persuasion, and Attitude Change," in A. G. Greenwald, T. C. Grock and T. M. Ostrom (eds.), Psychological Foundations of Attitudes, (New York: Academic Press, 1968).

B. Hayes-Roth, "Evolution of Cognitive Structures and Processes," Psychological Review, 84 (1977), 260-278.

E. J. Johnson and J. E. Russo, "The Organization of Product Information in Memory Identified by Recall Times," by H. Keith Hunt (ed.), Advances in Consumer Research, Vol. 5, Association for Consumer Research, forthcoming.

D. Klahr and J. G. Wallace, Cognitive Development, (New York: John Wiley and Sons, 1976).

H. E. Krugman, "The Impact of Television Advertising: Learning without Involvement," Public Opinion Quarterly, 29 (1965), 344-356.

R. E. Nesbitt and T. Decamp Wilson, "Telling More Than We Can Know: Verbal Reports on Mental processes," Psychological Review, 84 (1977), 231-259.

D. A. Norman and D. E. Rumelhart, Explorations in Cognition, (San Francisco: W. H. Freeman, 1975).

J. C. Olson, "Theories of Information Encoding and Storage: Implications for Consumer Research," in Andrew Mitchell (ed.), The Effect of Information on Consumer and Market Behavior, (Chicago: American Marketing Association), forthcoming.

R. G. Pachella, "The Interpretation of Reaction Time in Information-Processing Research," in Barry H. Kantowitz (ed.), Human Information Processing: Tutorials in Performance and Cognition, (Hillsdale, NJ: Lawrence Erlbaum Associates, 1974).

R. W. Pew, "The Speed-Accuracy Operating Characteristics," Attention and Performance II, Acta Psychologica, 30 (1969), 16-26.

M. R. Quillian, "The Teachable Language Comprehender: A Simulated Program and Theory of Language," Communications of ACM, 12 (1969), 459-476.

M. Ray, "Marketing Communication and the Hierarchy of Effects," Working Paper, Marketing Science Institute, Cambridge, Mass., November 1973.

S. Sternberg, "The Discovery of Processing Stages: Extensions of Donder's Method," Attention and Memory Performance II, Acta Psychologica, 30 (1969), 421-457.

S. Sternberg, "Memory Scanning: Mental Processes Revealed by Reaction Time Experiments," American Scientist, 57 (1969), 421-457.

E. Tulving, "Episodic and Semantic Memory," in E. Tulving and W. Donaldson (eds.), Organization of Memory, (New York: Academic Press, 1972), 385-387.

E. Tulving and G. H. Bower, "The Logic of Memory Representations,'' in Gordon H. Bower (ed.), The Psychology of Learning and Motivation, 8, (New York: Academic Press, 1974).

P. Wright, "The Cognitive Processes Mediating Acceptance of Advertising," Journal of Marketing Research, 10 (1973), 53-62.