An Introduction to Semantic Variables in Advertising Messages

Karen A. Berger, Pace University
Robert F. Gilmore, Baruch College, City University of New York
ABSTRACT - This article presents a new way to analyze the technical composition of language used in advertising messages. Three semantic variables, reports, inferences and judgments, known collectively as the "semantic trichotomy", are introduced with a review of the general semantics literature. A new conceptualization which adapts these notions for future experimental study is provided. Two preliminary studies provide the main focus of this paper to prove that this distinction is clearly discernible in consumers' minds, an essential step prior to further experimental study.
[ to cite ]:
Karen A. Berger and Robert F. Gilmore (1990) ,"An Introduction to Semantic Variables in Advertising Messages", in NA - Advances in Consumer Research Volume 17, eds. Marvin E. Goldberg, Gerald Gorn, and Richard W. Pollay, Provo, UT : Association for Consumer Research, Pages: 643-650.

Advances in Consumer Research Volume 17, 1990      Pages 643-650


Karen A. Berger, Pace University

Robert F. Gilmore, Baruch College, City University of New York


This article presents a new way to analyze the technical composition of language used in advertising messages. Three semantic variables, reports, inferences and judgments, known collectively as the "semantic trichotomy", are introduced with a review of the general semantics literature. A new conceptualization which adapts these notions for future experimental study is provided. Two preliminary studies provide the main focus of this paper to prove that this distinction is clearly discernible in consumers' minds, an essential step prior to further experimental study.


A review of marketing and communication literature suggests a need for the study of elements of the English language which contribute to the creation of persuasive messages. A semantic dimension, hereafter referred to as the "semantic trichotomy", is proposed as one systematic way of analyzing the technical composition of language.

This topic has been selected in light of the need for systematic ways of analyzing and constructing messages. Numerous researchers have pointed out that relatively little work has been done in analyzing the technical composition of advertising messages (Fishbein and Ajzen, 1981; Petty and Cacioppo, 1986). Issues, such as strong vs. weak arguments, or argument length, and other factors which influence the effectiveness of the message as constructed, have been widely addressed by persuasion researchers (Petty and Cacioppo, 1984). However, little attention has been paid to confronting the language itself and asking the question, what elements of the language contribute to the creation of persuasive, or not so persuasive messages? (Petty and Cacioppo, 1986). For this reason, Fishbein and Ajzen (1981) have said that "the general neglect of the information contained in a message ... is probably the most serious problem in communication and persuasion research" (p.359).

At least three taxonomies superficially resemble the semantic trichotomy: the Fishbein and Ajzen (1981) distinction descriptive, inferential and informational beliefs, the Holbrook (1975, 1978) notion of factual and evaluative messages and the notion of concrete vs. abstract sentences (Begg and Paivio, 1969; Paivio 1971; and Yuille and Paivio, 1979). Each of these approaches can be shown to be quite different from the semantic trichotomy.

This paper presents the first, necessary step in the process of developing and exploring the effects of these three semantic variables. In the first section, the semantic trichotomy is introduced presenting definitions from the general semantics literature. A new conceptualization is then presented in order to 1) provide a way to clearly separate the semantic trichotomy for later experimental study and 2) explain a priori the process through which these variables operate. In sum, this conceptualization provides a way to incorporate the semantic trichotomy into a framework which will be used in later work to generate hypotheses for further study. The final sections present two preliminary studies which demonstrate that the semantic trichotomy is inherently discernible and differentiable.


The notions of reports, inferences, and judgments are presented in a well-known, classic general semantics book, Language in Thought and Action by S. I. Hayakawa (1972). The definitions which are given for these terms in the aforementioned book provide a starting point for developing a semantically oriented approach to persuasion.


Hayakawa (1972) defines a report as a statement about a stimulus based on sensory information, i.e. information which we or someone else has directly seen, heard, or felt. Furthermore, Hayakawa states that for a statement to be a report, it must be capable of verification, either directly or indirectly. Further, people often draw conclusions mistakenly when they confuse inferences and judgments for reports (Hayakawa, 1972; Korzybski, 1947).

Condon (1985) uses the term 'statement of description' to describe what Hayakawa calls a report. He argues that these types of statements are generally more believable. Condon also states that reports generally do not create debate. Rather, they are accepted easily (Condon, 1985).


Hayakawa (1972) defines an inference as a statement about matters which are not directly observed. Unlike a report which is based directly on the sensory information at hand, the observed sensory information is used as a starting point to draw conclusions. This process of drawing conclusions is called abstraction. There are numerous levels of abstraction. For example, one can reason that one's dentist bills will be lower if one uses a toothpaste which reduces plaque and retards tooth decay. Or one can reason that one will be more attractive to members of the opposite sex.

Condon (1985) emphasizes the lower verifiability of inferences as compared to reports. According to Condon, some inferences are immediately verifiable; others take longer to verify; still others are totally unverifiable. Rosenthal (1971) adds that if a message lacks verifiability, the pre-existing knowledge of the listener may be essential to interpret the message in terms of its "truth" or meaning.

This discourse concerning verifiability is very interesting, but an inference which cannot be verified is still an inference, eliminating this as a distinctive feature. Furthermore, if some inferences can be verified, then how does one tell the difference between a verifiable report and a verifiable inference?

Condon re-states the Korzybskian notion that an inference often has less truth value. This lower truth value is explained by Gorman (1962) in terms of the strength of the predictions about the future which can be made.


Hayakawa (1972) defines a judgment as an expression of the writer's evaluation, often including approval or disapproval, of the event, persons or things being described. He provides this example: The report statement "The Senator was the only one who voted against the bill" can be turned into a judgment with the sentence 'The Senator courageously stood by his principles and voted against the bill." The judgment has been made that the Senator has behaved in a morally superior fashion. In effect, the author is giving his stamp of approval. According to Condon (1985), a statement of judgment adds the values of the person making the judgment to the meaning of the statement.

Values do not provide the basis for a clear definition, since neither reports nor inferences include the person making the statement. Agreement does not distinguish a judgment from the others. Consensus can be extensive. For example, few would argue with the judgment "tall willowy blondes are beautiful".


The basic concepts of two theories, Kelley's attribution theory and Chaiken's heuristic model of persuasion, are used in developing the new conceptualization. While none of the attribution approaches works in its entirety to explain the semantic trichotomy, Kelley's explanation of causes of observable actions (Kelley, 1967, 1972a, 1972b) is the most useful approach for separating the semantic trichotomy so that operationalization can take place. Further, Kelley and Chaiken's work provide insight into the mechanisms which may be operative in the persuasion process. In the case of the semantic trichotomy, the "causes" of the statements are not actions, but rather underlying beliefs about the meaning of the information provided. These underlying beliefs communicated by the semantic trichotomy mediate the persuasion process. The cause for an event is seen as part of the individual or part of the environment. This basic notion can be applied to situations in which people are presented with messages and must interpret their meaning and assess their validity.

An important implication of the attribution framework is that persuasion involves message recipients engaged in logical, consistent cognitive processing. An alternative to the systematic approach to message learning has been proposed by Chaiken (1980, 1982) and another similar version has been developed by Petty and Cacioppo (1981,1984,1986). Chaiken's heuristic model and Petty and Cacioppo's peripheral route to persuasion suggest that some message processing involves relatively little cognitive effort.

The concepts of report, inference and judgment can be viewed as message content terms which provide a set of information to consumers upon which various attributions are made. The persuasion process which occurs can be described in this way. (Figure A)

When message recipients hear or see a report, a high level of attribution to the stimulus occurs. When message recipients hear or see a judgment, a high level of attribution to the person occurs. In the case of inferences, the level of attributions lies somewhere in between reports and judgments.

The reading of a report, inference or judgment invokes a causal schema based on the stimulus x person x abstraction data revealed in the statement, suggesting heuristic information processing. This triggers the corresponding set of beliefs based on the data pattern. The naive psychology of human beings provides us with the ability to make this distinction. Based upon a person's experience with the external environment, one learns the underlying difference in these concepts.

A statement which is pure report, excluding all inferences and judgment, deals with properties of the stimulus in question. This characteristic provides the distinctive feature that separates the semantic trichotomy. The more a statement describes the stimulus exclusively in terms of those characteristics which can be perceived by the senses, the more the statement is a "pure" report excluding inferences and judgments. The perceived veridicality of the statement is derived from the attribution process in which, first, the statement is attributed to the stimulus and is perceived to be part of external reality. This is followed by the perception of the statement or message as more believable. Thus, the message is accepted more readily.

A judgment, on the other hand, is a statement which contains the affective response of the individual expressing the message. A pure judgment is one which only contains statements of how positive or negative the product is. Once again, the naive psychology of the listener/viewer provides the individual with the sense that the message is the property of the person as opposed to the stimulus. Last, a statement of inference contains different levels of abstraction from the stimulus or person.

Two issues which remain are (1) the effect of source on the perception of the product in the message and (2) the presence of systematic or heuristic processing. These questions have been left for consideration in another paper.

The conceptualization discussed here was then used to develop messages for preliminary testing. While it was recognized that the real issue was whether the semantic trichotomy can potentially add to our knowledge of persuasion, it was first necessary to demonstrate that each member of the semantic trichotomy is perceived as different.




Two preliminary studies were conducted. It was reasoned that if the semantic variables could be identified in an unguided sorting task, then this would be an indication that the report, inference, and judgment messages could persuade respondents differently.

Since it was hypothesized that there may be a source x semantic interaction, source and unmentioned source cells were included in the two studies which follow. However, it is not the scope of this paper to report on this dimension.

In each study, subjects were given 5 1/2" x 8 1/2" cards in random order. The following elements were kept constant:

1) For each category, each message was based on the same three attributes judged by the experimenter to be the three most important attributes in the category.

2) Each message contained three sentences.

3) The number of words on each card was the same plus or minus five words.

Respondents were deliberately given very little information in order to see if the three semantic variables could be identified. Specifically, they were instructed to sort the cards on the basis of the number of types of statements they perceived. Furthermore, they were told that the number and size of the piles was completely up to them. No other information was provided.

In each preliminary study, chi-square analysis of the data was performed to deter nine whether respondents were able to perceive the differences in the semantic trichotomy. Expected frequencies were developed based on perfect recognition of the distinction as will be explained below. Using the hypothesis that the observed frequencies are equal to the expected frequencies, it was expected that the chi-square values would not be significant if respondents were able to correctly place the statements into three or four piles.

Generally, scientific studies seek to reject the null hypothesis. However, chi-square analysis provides a clear way to analyze this sorting study; consequently, the experimental hypothesis is that there will be little deviation from the expected pattern. Thus, if the semantic trichotomy is real in subjects' minds, then observed frequencies will be equal or close to equal to the expected frequencies resulting in no significance.

The observed frequencies used to compute the chi-squares were defined as 'the number of subjects placing a report message of that category into a report dominant pile'. The dominant semantic variable in each pile was defined as the 'variable with the majority of cards in the pile'. For example, if a pile contained five cards with three report messages, one inference message, and one judgment message, it was labelled a report pile.

First Preliminary Study

Thirty-one Mercy College students participated in this experiment: Fourteen received cards with the messages introduced by a high source and seventeen received cards with the messages introduced by no source at all. Two levels of inference, referred to as "high" and "low", were included in this preliminary study. Five consumer categories were used: toothpaste, bar soap, shampoo, disposable diapers, and floor cleaner. Two sets of expected frequencies were developed (See Tables 1 and 2).

Results. Subjects were unable to correctly place the cards into piles when the results were tallied using two levels of inference. The difficulty of sorting into four piles was due to the respondents' inability to perceive the difference between the low and high inference categories.

When the chi-square analysis was performed combining the low and high inference messages, respondents were generally able to perceive the differences in the semantic trichotomy. The differences between the observed and expected were not significant for all five categories for the nonsource data and not significant for four out of five for the source data. Thus, the null hypothesis was not rejected (See Tables 1 and 2).

In light of this first preliminary study, a number of changes in the stimulus materials were made. First, the two inference levels were eliminated for the second preliminary study; second, the floor cleaner and disposable diaper categories were replaced by categories which, according to the judgment of the experimenter, are more salient to the student population from which subjects were recruited.





The six categories chosen for the second preliminary study were toothpaste, bar soap, analgesic, shampoo, cake mix and dog food. Improvements were made in the operationalization and a low credibility source condition was added.

Second Preliminary Study

Sixty-six students from Mercy College and The King's College were used to conduct the sorting exercise. Subjects for each of the three source conditions received a stack of eighteen cards containing one report, one inference, and one judgment message for each of the six categories. The procedure used in the sorting exercise was the same as the first preliminary study. Messages for one category are shown in Table 3. Expected frequencies appear in Table 4.

Results. Chi-square analysis revealed that almost all subjects were able to sort the cards into three categories. In all three source conditions, the observed frequencies were very close to the expected frequencies. The chi-square values were below one in the low and unmentioned source conditions, i.e. no statistically significant differences were found between the expected and the observed as predicted. The chi-square value in the high source condition was slightly higher (X2 =1.71), but still not statistically significant (See Table 4).


These two preliminary studies show that the distinction between report, inference and judgment is real in subjects' minds, although the terminology is not. Further study will be needed to deal with the larger issue of whether the notions of reports, inferences and judgments can potentially add to our knowledge of the persuasion process in advertising. It is hypothesized that the semantic trichotomy impacts on subjects' perception of goods and services and, hence, their intent to buy the products they have seen advertised. Since the preliminary studies suggest differences in the presence of source, this issue will also be incorporated into the model and will be further studied.






Begg, Ian and Allan Paivio (1969), "Concreteness and Imagery in Sentence Meaning," Journal of Verbal Learning and Verbal Behavior, 8, 821-827.

Chaiken, Shelly (1980), "Heuristic vs. Systematic Information Processing and the Use of Source Vs. Message Cues in Persuasion", Journal of Personality and 'Social Psychology, 5, 752-766.

Chaiken, Shelly (1982), "Heuristic Systematic Processing Distinction in Persuasion", paper presented at Symposium on Automatic Processing.

Condon, John C. (1985), Semantics and Communication, MacMillan Publishing Company, New York.

Eagly, Alice H., Shelly Chaiken, and Wendy Wood (1982), "An Attribution Analysis of Persuasion in New Directions in Attribution Theory and Research", in Attribution Theory and Research, Vol.3, J.H. Harvey and R.F.Kidd (Eds.), Lawrence Erlbaum Associates, Hillsdale, N.J., 37-62.

Fishbein, Martin and Icek Ajzen (1975), Belief, Attitude, Intention and Behavior: An Introduction to Theory and Research, Addison-Wesley Publishing Co., Reading, Mass., 132134.

Fishbein, Martin and Icek Ajzen (1978), "Acceptance, Yielding and Impact: Cognitive Processes in Persuasion," in Richard E. Petty, Thomas M. Ostrom and Timothy C. Brock (eds.),Cognitive Responses in Persuasion, Hillsdale, N.J., Lawrence Erlbaum Associates.

Gorman, Margaret (1962), General Semantics and Contemporary Thomism Lincoln, Neb., University of Nebraska Press.

Hayakawa, S.I. (1972), Language in Thought and Action. Harcourt, Brace, Jovanovich, Inc.

Holbrook, Morris B. (1975), A Study of Communication in Advertising, unpublished doctoral dissertation, Columbia University.

Holbrook, Morris B. (1978), "Beyond Attitude Structure: Toward the Informational Determinants of Attitude", Journal of Marketing Research, November, 15, 545-556.

Kelley, Harold H. (1967), "Attribution Theory in Social Psychology" in Nebraska Symposium on Motivation, David Levine (Ed.), University of Nebraska Press, Lincoln, Nebr.,192-240.

Kelley, Harold H. (1972a), "Attribution in Social Interaction" in Attribution: Perceiving the Causes of Behavior, Edward E. Jones, et al. (Eds.), General Learning Press, Morristown, N.J., 1-25.

Kelley, Harold H. (1972b), "Causal Schemata and the Attribution Process" in Attribution: Perceiving the Causes of Behavior, Edward E. Jones, et. al. (Eds.), General Learning Press, Morristown, N.J.

Korzybski, Alfred (1947), Science and Sanity: An Introduction to Non-Aristotelian Systems and General Semantics. Lancaster, Pa., Science Press Printing Co.

Paivio, Allan (1971), Imagery and Verbal Processes, Holt, Rinehart and Winston, New York, 59.

Petty, Richard E. and John T. Cacioppo (1981), Attitudes and Persuasion: Classic and Contemporary Approaches, Dubuque, Ia., William C. Brown Co.

Petty, Richard E. and John T. Cacioppo (1984), 'The Effects of Involvement on Responses to Argument Quantity and Quality: Central and Peripheral Routes to Persuasion," Journal of Personality and Social Psychology, 41, 69-81.

Petty, Richard E. and John T. Cacioppo (1986), "The Elaboration Likelihood Model of Persuasion," in L. Berkowitz (Ed.), Advances in Experimental Social Psychology, 19, Academic Press, New York.

Rosenthal, Paul I. (1971), "Specificity, Verifiability, and Message Credibility", Quarterly Journal of Speech, Dec., Vol. 57, -No. 4.

Yuille, John C. and Allan Paivio (1979), "Abstractness and Recall of Connected Discourse," Journal of Experimental Psychology, 82, 3, 467-471.