Professional Documents
Culture Documents
Assessing The Usefulness of Taxonomies (Comparison)
Assessing The Usefulness of Taxonomies (Comparison)
To cite this article: Bruce A. Huhmann & Pia A. Albinsson (2018): Assessing the Usefulness of
Taxonomies of Visual Rhetorical Figures, Journal of Current Issues & Research in Advertising, DOI:
10.1080/10641734.2018.1503106
Article views: 4
ABSTRACT
We compare six taxonomies (including some not previously tested) that
systematically categorize visual rhetorical figures (RFs) on their domain
coverage, reliability, parsimony, and predictive validity to determine which
is best for advertisers and researchers. All taxonomies showed increased
effectiveness with RFs than without RFs in a sample of 952 copy-tested
advertisements. However, one taxonomy—that of McQuarrie and Mick
—demonstrated the best balance of the criteria for a useful taxonomy
including predictive validity. Ads containing RFs from each of their visual
RF categories improved ad effectiveness measures related to Starch Noted,
Associated, and Read Most scores as predicted by theory.
Over the past two decades, researchers have been increasingly interested in how advertising rhet-
oric influences consumer processing of ad information (for reviews, see Huhmann 2007; Messaris
1997; Mohanty and Ratneshwar 2016; van Mulken, van Hooft, and Nederstigt 2014). This has
mostly focused on one aspect of rhetoric—rhetorical figures (RFs). Although some (e.g., Maes
and Schilperoord 2008) critique this definition, RFs have commonly been defined in the advertis-
ing literature as artful deviations from audience expectations with the goal of communicating
more than is literally expressed (e.g., McQuarrie and Mick 1996, 1999; Mothersbaugh, Huhmann,
and Franke 2002; Phillips and McQuarrie 2002; Theodorakis, Koritos and Stathakopoulos 2015).
The deviations are from expectations established by conventions in advertising, popular culture,
or personal experience and can include surface-level deviation in structure or arrangement of for-
mal image characteristics or semantic-level deviation in initially ambiguous or multiple meanings
(Callister and Stern 2007; Huhmann 2007).
Verbal RFs appear in about 70% of print ads (Leigh 1994; McQuarrie and Mick 1996). Visual
RFs are also common, but prevalence varies widely depending on the visual RF taxonomy
applied. For example, van Mulken’s (2003) content analysis of magazine ads found 86% contained
visual RFs when applying McQuarrie and Mick’s (1996) taxonomy versus 32% with Groupe Mu’s
(1980; 1992) taxonomy. These vast differences highlight one need for determining the best tax-
onomy for categorizing visual RFs.
While theoretical work from classical rhetoric can and has been applied to verbal RFs in
advertising (e.g., Leigh 1994; McGuire 2000; McQuarrie and Mick 1996), classical rhetoric
developed for verbal communication is not always suitable or applicable to visual communication
(Hill 2004). Scott (1994) called for developing visual rhetoric theories. She posited that ad images
could be seen as a form of discourse capable of nuance, subtlety, and the ability to abstract an
object’s essential characteristics. Although theoretical work on visual RFs was previously sparse,
researchers from advertising, consumer research, and other disciplines have begun developing
theories of visual RFs’ forms and effects. Advertising is a rich source for developing visual rhet-
oric theories due to its image-based arguments that stand alone, mirror, extend, or subvert verbal
communication (Messaris 2009). In developing theories of visual RFs, researchers have proposed
several taxonomies. Some adapted verbal RF taxonomies to the visual modality (e.g., McQuarrie
and Mick 1999). Others developed taxonomies specific to visual communication’s distinct charac-
teristics (e.g., Durand 1970; 1987: Forceville 1996; 2007; Gkiouzepas and Hogg 2011; Groupe Mu
1980; 1992; Phillips and McQuarrie 2004). Some of these taxonomies focus solely on semantic-
level deviations in which visual RFs imply meanings that must be inferred by substitution, discov-
ering connections, or appreciating multiple interpretations. Other taxonomies include visual RFs
with both semantic-level and surface-level deviation in which arrangements or patterns of image
elements or characteristics reinforce a particular meaning or attract attention to the message.
This study’s purpose is to compare six competing visual RF taxonomies to determine
which offers the best combination of Hunt’s (1991) characteristics of a useful taxonomy and to
recommend which taxonomy best serves the needs of advertising researchers and practitioners.
A taxonomy is most useful when it has (1) parsimony to avoid needlessly complex structure or
overabundant categories, (2) collectively exhaustive categories that cover a domain, (3) reliability
in application due to clear definitions and mutually exclusive categories, and (4) predictive valid-
ity in explaining differences consistent with theory (Hunt 1991). Theory and prior research hold
that advertising rhetoric should be associated with enhanced attention and processing (e.g.,
Huhmann 2007; McQuarrie and Mick 1996, 1999; Mothersbaugh et al. 2002). Thus, predictive
validity will first be assessed as the degree to which ads categorized as possessing visual RFs
within each taxonomy outperform those considered as not possessing visual RFs on advertising
effectiveness measures. A second aspect of predictive validity will be to determine whether differ-
ences in advertising effectiveness associated with a taxonomy’s categories confirm theoretical dif-
ferences proposed for that taxonomy. A taxonomy’s ability to classify visual RFs that benefit ad
attention and processing will make that taxonomy more useful to (1) practitioners or students of
advertising who wish to create new visual RFs for advertisements and (2) researchers who wish
to identify impactful visual RFs for further study or code instances of visual RFs within a sample
of ads or other images. Finally, this study will determine whether taxonomies covering both sur-
face-level and semantic-level deviation or only semantic-level deviation best account for observed
differences in effectiveness due to visual RFs in advertising. The study contributes to the advertis-
ing rhetoric literature as it offers a comparison of visual RF taxonomies’ effect on consumer proc-
essing of actual ads under natural rather than laboratory exposure conditions via Starch
Readership scores for a large number of ads.
Although researchers agree that RFs increase attention and processing (e.g., McQuarrie and
Mick 1999, 2009; Mothersbaugh et al. 2002), they disagree about what types of images should be
considered as visual RFs. For example, Phillips and McQuarrie (2004) limit visual RFs to images
with semantic-level deviation, which some researchers refer to as visual tropes (e.g., McQuarrie
and Mick 1999; Maes and Schilperoord 2009). These visual tropes have greater meaning openness
and can create one or more inferences or implicatures as consumers decode the exaggeration,
replacement, juxtaposition, exchange, or opposition of image elements that creates ambiguous or
multiple meanings (Huhmann 2007; Maes and Schilperoord 2009; Mothersbaugh et al. 2002). For
example, a study of one type of visual tropes in advertising found that consumers interpreted an
ad’s message by deciding among the various strong and weak implicatures (Phillips 1997).
Initially the meaning of a visual trope is ambiguous or polysemous and significant cognitive effort
must be applied to fill in a gap for successful interpretation. In other words, visual tropes’ seman-
tic-level deviation from realistic portrayals or pictorial conventions can trigger “thinking into” an
ad, whereby consumers use advertising literacy and prior brand or product-category knowledge
to winnow competing implicatures down to one or more interpretations that they believe the
advertiser intended (Phillips 1997). An example of visual tropes’ semantic-level deviation from
realistic portrayals or pictorial conventions is a detergent cup filled with a cloudy sky (McQuarrie
and Phillips 2005). Consumers must devote greater cognitive processing to decode this image
than they would with a more realistic or conventional one.
Alternatively, other researchers do not limit visual RFs to images with semantic-level deviation,
because advertisers can also create surface-level deviation in images’ formal characteristics (e.g.,
McQuarrie and Mick 1999; Maes and Schilperoord 2009). Visual RFs with surface-level deviations
focus attention on, reinforce, or emphasize a particular meaning or relationship through patterns
or arrangements of objects, shapes, angles, colors, textures, positions, gestures, spatial orientation,
or other formal image characteristics. For example, an Air France ad repeats the image of an air-
plane more than 50 times to attract attention to the ad and to emphasize the meaning that Air
France has a large number of flights. Although the same meaning is evident in the headline
“Serves More Cities in More Countries Than Any Other Airline,” the repetition of an identical
visual element heightens the focus on and reinforces this message. Some researchers refer to
images with surface-level deviation as visual schemes (e.g., Huhmann 2007; McQuarrie and Mick
1999, Maes and Schilperoord 2009). Visual schemes’ surface-level deviation creates excessive regu-
larity in which more than enough information from formal characteristics is available to decipher
the intended meaning, which in semiotics is termed “overcoding” (for more information, see
Huhmann 2007; McQuarrie and Mick 1996, 1999).
To investigate whether visual RF taxonomies that focus solely on visual tropes or ones that
include both visual schemes and visual tropes better account for differences in consumer process-
ing of advertisements with or without visual RFs, this study compares and analyzes six visual RF
taxonomies. Two taxonomies include both visual schemes and visual tropes—ones developed by
Durand (1987) and McQuarrie and Mick (1999, 2003). Four other taxonomies identified in the
advertising literature focus solely on visual tropes—ones developed by Forceville (1996, 2007),
Gkiouzepas and Hogg (2011), Groupe Mu (1992), and Phillips and McQuarrie (2004). The first
research question asks:
RQ1: Do ads containing visual schemes improve consumer processing outcomes over ads with no RF?
visual element. user may be suppressed to give greater object covers most of the ad (exaggeration), (2) relative sizes of elements modified
importance to the product or the product a reduced object or, in extreme, a totally white (e.g., a small person next to a giant
may be suppressed; so, readers must fill it page (understatement), (3) color object replaces bottle) unlike SUB-ID in which the
in. the same object in a monochrome image, etc. size of the whole visual is modified.
ADD-SIM -- More than one similar SPR-SIM -- Something similar in form SUB-SIM -- An implicit comparison (A is like EXC-SIM -- Similarity of form
(not identical) visual or visual takes the place of the suppressed object. B) is made through substitutions based on form between an abstract and a concrete
Similarity
element: stages of making/using a For example, a dotted line draws the form (i.e., one element replaces another that is element (e.g., police officers ticketing
product; different uses of a product; of the missing element or part of an object similar in shape or color, such as a banana a woman for using a speedy cleaning
or different people with the same is missing, but the missing part can be replacing the receiver of a yellow telephone) or solution) or content/meaning in
product or attitude/expression in the filled in because the first object is linked content/meaning (e.g., a visual figuration of an successive images (e.g., raw steak
same position. with or similar to another object, such as a abstract propriety, such as a parachute attached with an instant potato packet, then
mirror showing the missing part. to a car to illustrate safety). cooked steak with mashed potatoes.)
ADD-DIF -- The aim is to give the SPR-DIF -- Suspense or digression by SUB-DIF -- An element replaces another that is EXC-DIF -- An image fragmented
Difference
impression of a great number, delaying the presentation of one element physically connected or related to it. For into parts (e.g., cutting or tearing a
density, disparity, or diversity: by instead presenting some other elements example, a shoe can be replaced by its print, a picture or a fragmented face with
different people using different unrelated to it. This technique is more person by a hand, the law by a judge, a car by a lipstick being applied to lips and eye
products in different forms (unlike common in multiple page ads. wheel, wool by a sheep, or a product by its shadow to eyes, but the rest of the
ADD-SIM) or an accumulation of feature. face missing).
different products.
ADD-OPP -- Comparing opposing SPR-OPP -- Two elements (e.g., SUB-OPP -- An element replaces another to EXC-OPP -- (1) Impossible or
elements or products is a visual competing products) are placed in create a roundabout way of indicating unlikely images made by a photo-
Opposition
antithesis (e.g., “smiling” versus opposition but, one element of this something that may be too blunt, harsh, montage or computer alteration (e.g.,
“frowning” croissants). Separate opposition is not shown. sensitive, or private to actually show (e.g., someone bathing in a field of
similar pictures compare two flower petals may replace a tampon to indicate flowers) or (2) an unlikely exchange
products (e.g., a before and after comfort and freshness). Or instead of showing of elements (e.g., father and son
situation, two types of users, or two both a before and an after picture, an ad may exchanged magazines or English fox
extreme product varieties). more suggestively show just the before. hunters wearing motorcycle jackets).
ADD-FH -- The same visual SPR-FH -- an element seems hidden, but is SUB-FH -- Substituting something that initially EXC-FH -- Two apparently opposite
False Homology
element is shown twice, with really visible: (1) part of an element is seems unrelated or conveys the contrary to elements or elements with contrasting
different meanings or interpretations, covered, but it is still obvious what it is, what is shown (e.g., a tin can impossibly meaning are exchanged (e.g., a girl
or two different visuals elicit the (2) false modesty (e.g., a nude crosses her broken like a bottle to show metal’s solidity). with a swimsuit in the snow, not the
same meaning. arms in front of her breasts or her hands sand. Persons or products shown
cover her eyes), or (3) a product is only both right way up and upside-down).
half seen, in a silhouette, or an outline.
agency. Durand (1970, 1978, 1987) classified visual RFs based on two criteria, rhetorical operation
(i.e., addition, suppression, substitution, and exchange) and relation between the variable elements
(i.e., identity, similarity, difference, opposition, and false homology). Rhetorical operations create
families of visual RFs that differ by element relation. One family is derived from the first
operation—addition. Addition presents multiple visual elements. The second operation—suppres-
sion—involves missing or suppressed elements that consumers must identify to understand the
intended message. Advertisers swap one visual for another when using the third operation, substi-
tution, or they modify element relationships with the fourth operation, exchange (see Figure 1).
Durand’s taxonomy categorized visual RFs based on thousands of ads that he had gathered and
his experience working in the French advertising industry. Later, Dyer (1982) applied Durand’s
taxonomy to her analysis of visual RFs in British advertisements.
Scheme Trope
(overcoding) (undercoding)
Figure 2. McQuarrie and Mick’s (1996, 1999, 2003) taxonomy of visual rhetorical figures in advertising.
tropes). A visual destabilization trope in a toothpaste ad shows a string of pearls in the shape
of a smile (Phillips 2000). Consumers are expected to interpret this as meaning that the tooth-
paste will make one’s teeth pearly white.
Two content analyses have applied McQuarrie and Mick’s taxonomy to code visual RFs in
advertising (Phillips and McQuarrie 2002; van Mulken 2003). Also, visual RFs according to their
taxonomy increased ad liking and elaboration over ads that were classified as not containing
visual RFs (McQuarrie and Mick 1999). Mothersbaugh et al. (2002) confirmed the theorized ad
processing effects in a series of studies with verbal RFs categories from McQuarrie and Mick’s
taxonomy. Specifically, they confirmed that tropes > schemes > no RF on extent of ad processing
in (1) experiments and (2) an archival study. The Mothersbaugh et al. (2002) experiments found
that RFs, especially tropes, elicited higher ratings of interest in processing ad copy than no RFs.
Also, they found larger differences in responses to weak versus strong argument copy for ads
headlines containing tropes than schemes than no RFs. They measured these responses in terms
of differences in brand attitudes as well as difference in the number of positive versus negative
message thoughts. Their archival study of copy-testing scores for actual ads found higher Starch
Read Most scores for ads containing tropes than for schemes than for no RFs.
Construed Conjunction (Elements can be in one or the same place) Disjunction (Elements are next to each other)
(intended) level
Clarification: A substitution in which one entity replaces Clarification: Only one entity in a comparison is shown,
another. but the other may be inferred from the entity shown.
Example: Cotton cosmetic removal pads have been Example: An automobile ad that visualizes how a car
replaced by scourers, to illustrate the cleaning properties may give you the impression that you are in three places
of the latter. at once, as is illustrated by a city map with three dots
indicating “you are here” without showing the car.
Note: The original descriptions derive from Groupe Mu (1992); these descriptions and examples appear in van Mulken (2003). The
clarification was developed by the authors to help coders in the current study better understand the categories.
Figure 3. Groupe Mu’s (1992) visual rhetoric taxonomy.
cognitive effort (Groupe Mu 1980, 1992). They contend that every image contains both a “plastic”
and an “iconic” level. The “plastic” level involves the colors, textures, shapes, and compositional
conventions used in the specific image. The “iconic” level contains the relationships and referen-
ces that allow meaning to be created. Their visual RF taxonomy operates at the iconic level,
wherein visual RFs involve a deviation that the sender creates but that is not beyond the target
audiences’ ability to decipher or interpret (Groupe Mu 1970, 1992; van Mulken 2003). Groupe
Mu’s (1992) taxonomy can be represented as a grid consisting of four cells with two columns
related to possible placements—either in one place if the RF images are fused or merged (i.e.,
conjunction) or adjacent/nearby if the RF images appear side by side (i.e., disjunction). The rows
represent visual RF in which all images in the RF are present (In Praesentia) or those in which
one is absent (In Absentia). Figure 3 shows the four cells: In Praesentia Conjoint (IPC),
In Praesentia Disjoint (IPD), In Absentia Conjoint (IAC), and In Absentia Disjoint (IAD). Similar
to McQuarrie and Mick’s (1999) taxonomy, Groupe Mu’s grid is categorized based on increasing
amount of cognitive effort from “degre zero” in which there is no RF, increasing to IPC to IPD
to IAC to IAD (Groupe Mu 1992; van Mulken 2003). Originally developed for semiotic analysis
of visual rhetoric in general, Maes and Schilperoord (2008) and van Mulken (2003) have applied
Groupe Mu’s taxonomy to visual RFs in advertising contexts.
Hybrid metaphors (MP2) compare two objects that are both (at least partially) pictured. The objects are
often combined or fused together into a hybrid object.
Pictorial similes, like hybrid metaphors, show two objects; however, the objects are not fused or combined
with pictorial similes but presented separately.
Verbo-pictorial metaphors rely on comparisons between a visual object and a verbal object (a word or
several words). The visual object is often an image of the advertised product. The verbal object must
be present to elicit the metaphorical meaning.
Meaning operation
Visual structure (‘A is associated with B’) (‘A is like B’) (‘A is not like B’)
Juxtaposition CON-JUX -- Two side- SIM-JUX -- Two images of two OPP-JUX -- Two pictures shown side-by-
(Two side-by- by-side images. different products are pictured side but that clearly illustrate the opposite
side images) side-by side in order for the qualities or attributes of the product.
consumer to find similarities
between them.
Fusion CON-FUS -- Two SIM-FUS -- Two images are OPP-FUS -- A picture showing two sides of
(Two combined combined images or fused together in order to bring a person or object fused together in order for
images) products. out general attributes of the consumer to identify with one of the sides,
images to one product. the better one, if using the product.
Replacement CON-REP -- Image SIM-REP -- Only one of two OPP-REP -- A present image is shown that is
(Image present present points to an images that are to be compared not like an absent image. A negative
points to an absent image. is shown in the ad. Some context inference is evoked by opposition but in the
absent image) is given to assist consumers to context of the positive emotions that attend
construct the missing image. the successful resolution of an artful
deviation. A rebuttal analogy is commonly
used (e.g., irony is useful in communicating
ridicule, condemnation, criticism,
impoliteness.)
Realistic symbiosis (no intruder) RS-JUX -- Unexpected similarities in color, RS-SYNTH -- Elements (e.g., color, texture,
position, or angle of view shared by two position, angle of view) of an unpictured
separate objects in an otherwise realistic object are incorporated into another object
scenario points to metaphorical implications (e.g., a Heinz ketchup bottle is sliced like a
(e.g., a red glass Heinz ketchup bottle is tomato).
positioned adjacent to a tomato).
Replacement (intruder) REP-JUX -- An object shown with something REP-SYNTH -- An object that is otherwise
replacing it that would otherwise be foreign to foreign to the context has replaced an
the context (e.g., a Chrysler convertible has unpictured object that is suggested by other
replaced one of a row of poolside chaise contextual elements (e.g., a stylish air-
lounges or a marmalade jar has replaced one conditioning unit has replaced a painting on
orange in a crate of oranges). an easel or an orange with a marmalade label
has replaced a jar of marmalade).
Artificial symbiosis (not applicable) AS-JUX -- Objects that would realistically AS-SYNTH -- A combination object is
not be pictured in the same scene are boldly created by unrealistically fusing two objects
placed together. The artificiality may arise without regard to proper size, material,
from a lack of a realistic background, texture, or contextual differences (e.g., a
perspective, positions, or proper size tablet computer is shown with the handle of a
differences, or (e.g., a Saab automobile is razor attached to its back or an object is
shown with the reflection of a jet fighter or a presented that is half marmalade jar and half
Toshiba flat screen television is shown next to orange).
a razor blade that is the same size).
Other characteristics of good taxonomies impact the ability of researchers to reliably place
an item into the same category. First, a good taxonomy should clearly define each category.
Clear definitions increase the ability of different researchers to independently assign an item to
the same category. Second, a good taxonomy should have categories that are mutually exclusive;
otherwise an item could be classified into two or more categories. An item should not fit
into multiple categories at the same level. However, a taxonomy may have different levels, such
as McQuarrie and Mick’s taxonomy with both the figurative mode and rhetorical operation levels.
Unfortunately, many taxonomies in marketing fail to meet the mutually exclusive characteristic
(Hunt 1991). A lack of mutual exclusive categories makes that taxonomy’s application more
ambiguous. Because greater reliability implies better definitions and clearer divisions between cat-
egories, this study investigates the degree to which independent coders concur in their ability to
recognize and classify the visual RFs in a wide variety of actual advertisements.
RQ3: Do the visual RF taxonomies exhibit adequate reliability?
Hunt (1991) states that a taxonomy should be useful in its ability to account for observed dif-
ferences in a phenomenon. That is, it should solve some kind of problem better than alternatives
or account for differences in associations, antecedents, or outcomes better than alternatives.
Thus, not only should a taxonomy help identify different instances of a phenomenon, but its
categorization should have predictive power (Funk 1983; Weber 2012). Thus, the final characteris-
tic of a good taxonomy relates to its predictive validity. If the categories are based on theoretical
differences that should influence outcomes or if prior research documents different outcomes,
then category membership should predict such outcomes.
Prior research has found verbal RFs boost copy-testing measures of ad effectiveness
(McQuarrie and Phillips 2008; Mothersbaugh et al. 2002; Tom and Eves 1999). Also, experimental
10 B. A. HUHMANN AND P. A. ALBINSSON
research has found that ads with rather than without verbal or visual RFs improved ad effective-
ness outcomes similar to those measured via copy-testing, including attention (Huhmann and
Albinsson 2012), brand awareness (Gray and Snyder 1989; Huhmann and Albinsson 2012), extent
of processing (Mothersbaugh et al. 2002), and aided and unaided recall (Halkias and Kokkinaki
2014; McQuarrie and Mick 2003, 2009; Toncar and Munch 2001). Thus, to test predictive
validity, ads that a taxonomy classifies as possessing visual RFs should increase copy-testing
scores in comparison to those classified as possessing no visual RFs.
RQ4: Which taxonomy has the highest proportion of visual RF categories with ads that increase processing
over ads without visual RFs?
Developers of some visual RF taxonomies held that different dimensions or categories would
yield differences in cognitive effort or advertising effectiveness (e.g., Groupe Mu 1992; McQuarrie
and Mick 1996). Both Phillips and McQuarrie (2004) and Gkiouzepas and Hogg (2011) also
included dimensions theorized to specifically relate to consumer advertising processing. In fact,
Phillips and McQuarrie (2004) recognize categorization unrelated to differences in outcomes
as a taxonomy’s most vulnerable potential weakness. Although developers of each taxonomy or
those who have previously applied them may have predicted category differences based on
a variety of theories, the final research question addresses this second aspect of predictive validity—do
category differences support theoretically predicted differences?
RQ5: Does each visual RF taxonomy exhibit ad processing differences between categories that fit theoretical
predictions?
Method
Sample
Research questions were examined using a sample of 952 ads from a wide variety of magazines:
Barron’s (two issues), Bon Appetit, Business Week, Country Living, Ebony, Glamour (two issues),
GQ, Home, House Beautiful, Industry Week, Men’s Journal, Newsweek, Parents, Real Simple, Sports
Illustrated, and Sunset. This convenience sample consisted of all magazine ads with copy-testing
scores provided by a subsidiary of the GfK Group, MRI Starch. Begun in 1923 as Daniel Starch
and Staff, MRI Starch is a leading professional research service measuring print ad readership.
Its “Starch scores” have been used in many previous studies in the marketing literature (e.g.,
Finn 1988; Franke et al. 2004; Huhmann et al. 2012; McQuarrie and Phillips 2008; Mothersbaugh
et al. 2002). Starch scores for ads in a particular issue are based on personal interviews with
a representative sample of 100 to 200 of a magazine’s readers. Using a rotating starting position,
interviewers go through the issue and ask whether the reader “noted” the ad, “associated” it with
the advertised brand, and “read most” (at least half) of the ad copy. An ad’s Starch scores are the
proportions of magazine readers that respond positively to each of these questions.
Starch scores are a recognition-based method for assessing print ad readership. Prior research
has determined that Starch Noted scores are a good proxy for attention to an advertisement,
Associated scores are a good proxy for brand processing, and Read Most scores (i.e., the percent
of respondents who report having read 50% or more of the ad’s copy) are a reasonable proxy for
extent of ad message processing (e.g., Finn 1988).
definitions were included in the coder training materials. The authors independently analyzed
and pretested the coder training materials and procedures before training three graduate students
to code the ads.
Identifying and classifying rhetorical figures can be difficult for those who lack experience and
familiarity with advertising rhetoric. For example, in her content analysis, van Mulken (2003)
observed relatively low levels of intercoder reliability due to coder insecurity about how to classify
particular visual rhetorical works. Similarly, in his content analysis of verbal rhetorical figures,
Leigh (1994) also initially experienced low intercoder reliability. To educate and familiarize the
three graduate student coders before coding the data for the current study, a three-step process
was used. First, the coders read through the definitions with the authors, who then provided a
number of illustrative examples for each category in each taxonomy. Then, the coders practiced
identifying and classifying visual RFs in all full-page or larger ads from a recent issue of Elle
magazine. Then, the authors discussed the disagreements in how a coder and authors coded the
presence or absence of each category in each ad. Fourth, coders again practiced identifying and
classifying visual RFs in all full-page or larger ads from a recent issue of Fitness magazine and
again discussed any disagreements with the authors. By the end of this process, coders were more
confident in applying the visual RF taxonomies to advertising and showed greater agreement in
coding how each taxonomy would classify an advertisement in terms of visual RFs.
Once training was complete, each coder worked independently to indicate the presence or
absence of every visual RF category from all taxonomies for each ad in the full sample.
Awareness of the presence of visual RFs was heightened through coding categories from all
taxonomies simultaneously as opposed to sequentially. More specifically, coders looked at one ad
at a time and recorded categorization data in Microsoft Excel for all six taxonomies before
moving on to the next ad. The scrutiny each ad image received during this simultaneous
application of taxonomies helped ensure a more complete identification of the visual RFs present
in the sample. After all coding was completed, disagreements were resolved by referring to visual
RF ad examples and definitions until agreement was reached for each ad. Intercoder reliability
levels for each taxonomy are compared in the “Analysis and Results” section.
Other variables
Variables found to account for much of the variance in Starch scores due to executional elements are
included as covariates in all the multivariate and univariate analyses—ad size, cover positions, front
versus back of book ad placement, color, visual size, copy length, and magazine vehicle in which the
ad appeared (Finn 1988, Franke et al. 2004, Huhmann et al. 2012; Mothersbaugh et al., 2002). These
executional characteristics were coded separately by two additional coders. Ad size was a three-
category variable (“less than one page”, “a full page,” or “more than one page”). Copy length was
a four-category variable ranging from “25 words or less” to “100 words or more.” Visual size was
a five-category variable ranging from “less than quarter of the ad” to “entire ad.” Magazine vehicle
was included to account for unique variance due to the ad processing effort that is devoted by
a particular magazine audience. Intercoder agreement on executional elements ranged from 78% for
visual size to 99% for cover position, with visual size being the only variable dropping below 89%
agreement. Disagreements were resolved by a third coder. All covariates had significant effects on
Starch scores (p’s .05) except front versus back of the book on Read Most scores and color on any
Starch scores, which was likely due to the low proportion (6.4%) of black-and-white ads in the sample.
whether ads containing visual schemes improve consumer processing outcomes over ads
with no RF.
Ads categorized as containing visual schemes in two McQuarrie and Mick categories—repeti-
tion schemes and reversal schemes—had higher scores on all three Starch scores than ads without
visual RFs (see Table 1). Also, ads classified as containing visual schemes appearing in four
Addition categories of Durand’s taxonomy have higher Noted scores than ads without RFs. Those
in Durand’s Rhyme/Comparision (ADD-SIM) and Accumulation (ADD-DIF) categories also have
higher Associated and Read Most scores than ads without RFs. Ads in Durand’s Anachronism/
Antithesis (ADD-OPP) category, but not in his Repetition (ADD-ID) category, also have higher
Associated scores than ads without RFs (see Table 2).
Table 1. Prevalence of ads and starch score means by visual rhetorical figure (RF) taxonomy.
Starch score means (standard errors)
Taxonomy
Category n PRL Noted Associated Read Most
Groupe Mu
No visual RFs 618 75 49.56a (0.43) 43.13a (0.48) 17.68a (0.35)
In Praesentia Conjoint (IPC) 88 93 53.55b (1.14) 46.80bd (1.26) 19.60bcd (0.99)
In Praesentia Disjoint (IPD) 171 63 55.39b (0.83) 49.46b (0.91) 20.57c (0.71)
In Absentia Conjoint (IAC) 33 100 60.23c (1.86) 56.17c (2.05) 21.57c (1.50)
In Absentia Disjoint (IAD) 42 99 50.61d (1.63) 43.80d (1.80) 17.00d (1.38)
Multivariate Wilks’s lambda ¼ .915; eta ¼ .029; F ¼ 6.560, p .001
2
Table 2. Prevalence of ads and starch score means for Durand’s visual RF taxonomy.
Starch score means (standard errors)
Rhetorical operation
relation category N PRL Noted Associated Read Most
No visual RFs 166 79 43.83a (0.85) 37.57a (0.94) 15.22a (0.71)
Additive
Repetition (ADD-ID) 18 85 50.70bc (2.46) 42.31bc (2.73) 16.89ab (1.99)
Rhyme/COmparision (ADD-SIM) 157 86 51.66bde (0.84) 45.55bde (0.93) 18.46bc (0.69)
Accumulation (ADD-DIF) 48 94 54.80cfgh (1.52) 48.52dfgh (1.68) 18.40bc (1.23)
Anachronism/antithesis (ADD-OPP) 45 96 51.11bfg (1.56) 44.06be (1.73) 17.94bc (1.28)
Antanaclasis/paradox (ADD-FH) 19 98 52.17cdef (2.39) 46.40bef (2.65) 18.62bcd (1.99)
Suppression
Ellipsis (SPR-ID) 13 96 52.48cdfi (2.89) 44.90bcf (3.20) 18.41acd (2.66)
Circumlocution (SPR-SIM) 6 99 53.40bcfh (4.26) 47.53befg (4.72) 18.24acde (3.45)
Suspense (SPR-DIF) 16 98 51.77abei (2.61) 41.29ac (2.98) 15.38ac (2.44)
Dubitation/reticence (SPR-OPP) 10 98 57.75cdfh (3.29) 47.65befg (3.65) 22.54bde (2.81)
Tautology/preterition (SPR-FH) 28 97 53.17bc (1.97) 46.13bef (2.19) 16.81acf (1.63)
Substitution
Homeophore (SUB-ID) 78 88 53.09bc (1.18) 47.20bf (1.32) 21.52de (1.00)
Allusion/metaphor (SUB-SIM) 49 92 52.87bc (1.49) 45.32bde (1.65) 18.87bcd (1.26)
Metonymy (SUB-DIF) 49 93 55.24ch (1.50) 49.95fg (1.66) 21.58de (1.24)
Periphrasis/euphemism (SUB-OPP) 10 98 49.77abe (3.30) 42.88ace (3.66) 18.47acd (2.99)
Pun/antiphrasis (SUB-FH) 16 98 55.06cdfhi (2.60) 48.70bfg (2.89) 24.54e (2.11)
Exchange
Inversion (EXC-ID) 75 89 54.00cdfi (1.22) 48.62fg (1.35) 20.21bd (1.06)
Hendiadys (EXC-SIM) 32 96 58.29h (1.85) 52.77g (2.05) 19.96bdf (1.50)
Asyndeton (EXC-DIF) 33 97 51.85bc (1.82) 46.47bef (2.02) 18.67bcd (1.52)
Anacoluthon/chiasmus (EXC-OPP) 73 96 51.81bf (1.23) 45.79eh (1.36) 18.03bc (1.06)
Antimetabole/antilogy (EXC-FH) 11 100 52.65bcfh (3.14) 46.55befg (3.48) 16.91acd (2.67)
Multivariate Wilks’s lambda ¼ .818 eta2 ¼ .065; F ¼ 2.959, p .001
Univariate F values 5.80c 5.14c 2.65c
Note. Different superscripts indicate significant planned contrast between categories cell means.
Thus, it appears that including visual schemes in a visual RF taxonomy does capture
differences in Noted scores (attention) and sometimes Associated (brand processing) and Read
Most (message processing) scores between ads classified as incorporating visual schemes versus
those classified as not incorporating visual RFs.
Domain coverage
Guidelines for a good taxonomy require that categories be collectively exhaustive of the domain.
The second research question asked which visual RF taxonomy had the most exhaustive coverage.
Table 3 shows the domain coverage of the six taxonomies. Column 1 shows the number of ads that
fit that taxonomy’s categories for visual RFs. Column 2 lists the number of ads that do not fit into
the visual RF categories for each taxonomy and, therefore, would be considered as not containing
visual RFs according to that taxonomy. To assess domain coverage, the last three columns show the
number of ads that coders considered as having some sort of visual RF using the broad definition
of artful deviance from conventional visual presentations but did not fit into a particular
taxonomy’s visual RF categories. Column 3 shows the number of ads considered to contain visual
schemes that coders were unable to categorize using a particular taxonomy. Column 4 shows the
number of ads considered to contain unclassifiable visual tropes. Column 5 shows the total number
of ads in the visual RF domain that were unclassifiable using each taxonomy.
Visual RFs may involve artful deviation at the surface level through arrangements, designs, or
patterns of formal image characteristics (e.g., objects, colors, shapes, gestures, positions) to create
emphasis or focus message processing via visual schemes. Then again, they may involve artful
deviation at the semantic level to create multiple inferences, interpretations, or implicatures
14 B. A. HUHMANN AND P. A. ALBINSSON
through ambiguous or multiple meanings via visual tropes (Huhmann 2007; Mothersbaugh et al.
2002). As mentioned before, the taxonomies by Groupe Mu (1992), Forceville (1996), Phillips
and McQuarrie (2004), and Gkiouzepas and Hogg (2011) more narrowly focus on visual tropes.
Naturally, none of these four taxonomies accounted for the 223 ads that coders considered to
contain visual schemes. However, coders believed that each of these four taxonomies is also
unable to account for many instances of visual tropes. Of course, Forceville’s taxonomy has the
least domain coverage as it only concerns visual metaphors. Durand’s taxonomy is the most
exhaustive, because it accounts for the most instances of visual RFs overall, even when the
analysis is limited solely to visual tropes, closely followed by McQuarrie and Mick’s taxonomy.
Reliability
The third research question asked whether or not each visual RF taxonomy would exhibit
adequate reliability. Reliability is computed using Rust and Cooil’s (1994) Proportional Reduction
in Loss (PRL) technique. PRL involves the expected loss from erroneous coding decisions
and extends techniques of Cronbach’s alpha and Perreault and Leigh’s (1989) index of reliability
to coding decisions with three or more coders. As shown in Tables 1 and 2, all visual RF
taxonomies achieved adequate intercoder reliabilities for most categories.
However, some categories had intercoder reliability dropping below Nunnally’s (1978, 245)
recommendation of .70 or better for the reliability of instruments used in basic research. These
were Gkiouzepas and Hogg’s realistic symbiosis juxtaposition (PRL ¼ 54) and artificial symbiosis
juxtaposition (PRL ¼ 68) and Groupe Mu’s In Praesentia Disjoint (PRL ¼ 63). In Phillips and
McQuarrie’s taxonomy, no category dropped below a PRL of 71. In Forceville’s taxonomy,
no category dropped below a PRL of 79. In Durand’s taxonomy, no category dropped below
a PRL of 79. In McQuarrie and Mick’s taxonomy, no category dropped below a PRL of 88.
The taxonomy with the highest overall reliability average across all categories was McQuarrie and
Mick’s taxonomy (average PRL ¼ 95).
Predictive validity
Research question 4 asked about the degree to which each visual RF taxonomy exhibits predictive
validity in terms of ads coded as possessing visual RFs having higher Noted, Associated, and
Read Most scores than ads coded as not possessingt visual RFs. The three Starch scores were
significantly correlated (all r’s < 0.578, p’s < .001). Grand means for the 952 total ads were Noted
score (51.39), Associated score (45.10), and Read Most (18.45).
Multivariate analyses of covariance (MANCOVAs) with univariate analysis of covariance
follow-ups were performed to determine how ads classified via each taxonomy affected Starch
scores. Executional and mechanical variables—ad size, cover positions, front versus back of book
ad placement, color, visual size, copy length, and magazine vehicle in which the ad appeared—
JOURNAL OF CURRENT ISSUES & RESEARCH IN ADVERTISING 15
were included as covariates in all the multivariate and univariate analyses because prior research
has found these variables to account for considerable variance in Starch scores (Finn 1988;
Franke et al. 2004; Huhmann et al. 2012; Mothersbaugh et al. 2002).
The category n’s (i.e., the number of ads), Starch score marginal means adjusted for the
covariates, and the standard error for each taxonomy are presented in Tables 1 and 2. Standard
error (SE) is reported because it takes into account both the standard deviation (sd) and the
sample size (N), such that SE ¼ sd/冑N.
First, results are presented for those taxonomies without coverage of visual schemes. For the
Groupe Mu taxonomy, MANCOVA revealed differences across the combined dependent variables
between ads coded as without visual RFs (i.e., no category in Groupe Mu’s taxonomy), IPC, IPD,
IAC, and IAD (see Table 1). Univariate follow-ups indicated effects for each of the three individual
Starch scores as well (p’s .001). Planned comparisons show that ads in two of Groupe Mu’s cate-
gories (i.e., IPD and IAC) exhibit higher Noted, Associated, and Read Most scores than ads without
visual RFs. Ads in another category, IPC, exhibit higher Noted and Associated scores, but not Read
Most scores, than ads without visual RFs. However, ads with RFs categorized as IAD do not differ
in Noted, Associated, or Read Most score from ads without visual RFs (p’s .189).
For Forceville’s taxonomy, MANCOVA also revealed differences across the combined
dependent variables between ads categorized as without visual RFs, contextual metaphors, hybrid
metaphors, pictorial similes, and verbo-pictorial metaphors (see Table 1). Univariate follow-ups
indicated effects for each of the three individual Starch scores as well (p’s .001). Planned
comparisons show that ads in three of Forceville’s categories (i.e., contextual metaphor, pictorial
simile, and verbo-pictorial metaphor) exhibit higher Noted, Associated, and Read Most scores
than ads without visual RFs. Ads in another category, hybrid metaphor, exhibit higher Noted and
Associated scores, but not Read Most scores, than ads without visual RFs.
For Gkiouzepas and Hogg’s taxonomy, MANCOVA also revealed differences across the com-
bined dependent variables between the categories (see Table 1). Univariate follow-ups indicated
effects for each of the three individual Starch scores as well (p’s .001). Planned comparisons
show that ads in three of Gkiouzepas and Hogg’s categories (i.e., realistic symbiosis synthesis,
replacement synthesis, and artificial symbiosis juxtaposition) exhibit higher Noted, Associated,
and Read Most scores than ads without visual RFs (see Table 1). Ads in two of Gkiouzepas and
Hogg’s other categories (i.e., realistic symbiosis juxtaposition and artificial symbiosis synthesis)
exhibit higher Noted and Associated scores, but not Read Most scores, than ads without visual
RFs. Finally, no difference in Starch scores is observed between ads in Gkiouzepas and Hogg’s
replacement juxtaposition category and ads without visual RFs.
For Phillip and McQuarrie’s taxonomy, MANCOVA revealed differences across the combined
dependent variables between the categories. Univariate follow-ups indicated overall effects for
each of the three individual Starch scores as well (p’s .001). Planned comparisons show that ads
in three Phillip and McQuarrie categories (i.e., CON-JUX, CON-FUS, and SIM-REP) exhibit
higher Noted, Associated, and Read Most scores than ads without visual RFs (see Table 1).
Ads in another category, SIM-JUX, exhibit higher Noted and Associated scores, but not Read
Most scores, than ads without visual RFs. Ads in five of their categories show no Starch score
increases over ads without visual RFs (p’s .099).
Part of the difficulty may be the small number of ads within some Phillips and McQuarrie
categories, because their taxonomy has nine categories compared to four categories for Groupe
Mu or Forceville or six for Gkiouzepas and Hogg’s taxonomy. Therefore, further MANCOVAs
with univariate follow-ups were conducted looking at ads classified by each Phillips and
McQuarrie dimension separately. First, examining differences between ads without visual RFs and
those classified within the three visual structures (i.e., juxtaposition, fusion, and replacement)
reveals an overall multivariate effect (Wilks’s lambda ¼ .947, eta2 ¼ .018, F ¼ 4.636, p .001)
and univariate effects on two Starch scores (Noted: F ¼ 10.908, p .001; Associated: F ¼ 11.746,
16 B. A. HUHMANN AND P. A. ALBINSSON
p .001; Read Most: F ¼ 2.477, ns). Planned comparisons of ads within each visual structure
dimension against ads without visual RF found that ads classified as juxtaposition exhibit higher
Noted, Associated, and Read Most scores (p’s .007), whereas ads classified as fusion and
replacement exhibit higher Noted and Associated (p .02), but not Read Most scores (p .408).
Second, examining differences between ads without visual RFs and Phillips and McQuarrie’s
three meaning operation dimensions (i.e., connection, comparison for similarity, and comparison
for opposition) reveals an overall multivariate effect (Wilks’s lambda ¼ .945, eta2 ¼ .019,
F ¼ 4.858, p .001) and univariate effects on each Starch score (Noted: F ¼ 12.542, p .001;
Associated: F ¼ 13.251, p .001; Read Most: F ¼ 2.621, p .05). Planned comparisons of ads
within each meaning operation dimension against ads without visual RFs reveal that ads classified
as connection exhibit higher Noted, Associated, and Read Most scores (p’s .007) and ads classi-
fied as similarity exhibit higher Noted and Associated (p’s .001), but not Read Most scores
(p ¼ .187); however, Starch scores for ads classified as opposition do not differ from those for ads
without visual RFs (p’s .144). One caveat is that relatively few instances (N ¼ 65) of opposition
were observed.
Next, results are presented for the two taxonomies that include both visual schemes and visual
tropes. For McQuarrie and Mick’s taxonomy, MANCOVA revealed differences across the com-
bined dependent variables. Also, univariate follow-ups indicated effects for all three individual
Starch scores (p’s .001). Planned comparisons show that ads within three of McQuarrie and
Mick’s categories (i.e., repetition schemes, substitution tropes, and destabilization tropes) exhibit
higher Noted, Associated, and Read Most scores than ads without visual RFs (see Table 1). Ads
within the other category, reversal schemes, exhibit higher Noted and Associated scores, but not
Read Most scores, than ads without visual RFs. As in van Mulken’s (2003) content analysis of vis-
ual RFs in a sample of French and Dutch ads, destabilization tropes were the most commonly
identified category in the current sample, followed by substitution tropes, then repetition schemes
with reversal schemes as the least commonly identified category (N ¼ 39).
For Durand’s taxonomy, MANCOVA revealed differences across the combined dependent
variables. Univariate follow-ups indicated effects for each of the three individual Starch scores
(p’s .001). Planned comparisons show that ads within nine of Durand’s categories (i.e., ADD-SIM,
ADD-DIF, SUB-ID, SUB-DIF, SUB-FH, EXC-ID, EXC-SIM, EXC-DIF, and EXC-OPP) exhibit
higher Noted, Associated, and Read Most scores than ads without visual RFs (see Table 2). Ads
within another seven categories (ADD-OPP, ADD-FH, SPR-SIM, SPR-OPP, SPR-FH, SUB-SIM,
and EXC-FH) exhibit higher Noted and Associated scores, but not Read Most scores, than ads
without visual RFs. Ads within two visual rhetoric categories (ADD-ID and SPR-ID) increased
Noted scores only compared to ads without visual RFs. Finally, ads within two other categories
(SPR-DIF and SUB-OPP) show no Starch score increases over ads without visual RFs (p’s .113).
Due to the large number of RF categories in Durand’s taxonomy, it was difficult to identify
sufficient numbers of ads for each category even within our large sample. Therefore, further
MANCOVAs with univariate follow-ups were conducted looking at ads with RFs categorized by
each of Durand’s dimension separately. First, examining differences between ads without visual
RFs and ads classified within each of the four rhetorical operations (i.e., addition, suppression,
substitution and exchange) reveals an overall multivariate effect (Wilks’s lambda ¼ .862,
eta2 ¼ .048, F ¼ 9.729, p .001) and univariate effects on each Starch score (Noted: F ¼ 23.368,
p .001; Associated: F ¼ 19.534, p .001; Read Most: F ¼ 9.611, p .001). Planned comparisons
reveal that ads within each rhetorical operation exhibit higher Noted, Associated, and Read Most
scores than ads without visual RFs (p’s .009).
Second, examining differences between ads without visual RFs and ads classified by Durand’s
five element relations (i.e., identity, similarity, difference, opposition, and false homology) reveals
an overall multivariate effect (Wilks’s lambda ¼ .881, eta2 ¼ .041, F ¼ 6.581, p .001) and univari-
ate effects on each Starch score (Noted: F ¼ 19.281, p .001; Associated: F ¼ 15.409, p .001;
JOURNAL OF CURRENT ISSUES & RESEARCH IN ADVERTISING 17
Read Most: F ¼ 5.707, p .001). Planned comparisons reveal that ads within each element relation
exhibit higher Noted, Associated, and Read Most scores than ads without visual RFs (p’s .01).
To address RQ5, ads within each category of a taxonomy were compared. Then we observed
whether or not actual differences fit theoretical explanations.
Once again, investigation into RQ5 start with the taxonomies focused solely on visual tropes.
For Groupe Mu’s taxonomy, cognitive effort is predicted to increase from ads classified
as containing no visual RF to IPC to IPD to IAC to IAD (Groupe Mu 1992; van Mulken 2003).
As shown in Table 1, this prediction is partially supported as Starch scores increase from ads
without RFs to ads containing IPC or IPD, which have similar Starch scores, contrary to predic-
tion, to ads classified as containing IAC. However, ads classified as containing IAD have much
lower Starch scores, which are almost as low as ads without visual RFs. Actually, rather than
exhibiting increasing cognitive effort, ads containing RFs included in Groupe Mu’s taxonomy
appear to exhibit increasing cognitive load resulting in a nonmonotonic upside-down U-shaped
function between cognitive load and processing as predicted by some researchers for RFs
(e.g., Huhmann 2007; Mohanty and Ratneshwar 2016).
Forceville (1996, 2007) does not offer a prediction of processing-related differences between
his categories. In fact, ads classified using his categories show a great deal of overlap in their
levels of Starch scores. However, across Starch scores, ads classified as contextual metaphors
outperform ads classified as hybrid metaphors, with ads classified as pictorial similes and
verbo-pictorial metaphors bridging the gap between them.
Gkiouzepas and Hogg predicted and found that visual RFs in their synthesis mode of
representations would provoke greater elaboration than those in their juxtaposition mode of
representation with their six experimental ad stimuli. They also found that some visual RFs
within their synthesis mode of representation appeared to have a negative impact on the compre-
hension compared to those within their juxtaposition mode of representation. In the current
study, ads classified into their synthesis mode of representation outperformed their juxtaposition
counterparts for their replacement visual scenario RF categories (REP-SYNTH vs. REP-JUX).
Results were in the right direction but did not achieve the level of significance for their realistic
symbiosis visual scenario categories (RS-SYNTH vs. RS-JUX). With their artificial symbiosis
visual scenario categories, ads classified as AS-JUX achieved higher Associated and Read Most
scores than ads classified as AS-SYNTH, although Noted scores were similar.
Phillips and McQuarrie (2004) predict increasing cognitive elaboration from the top left to
the bottom right categories of their taxonomy (see Figure 4). However, this is not evident in
the current study’s Starch score results. Again, increasing cognitive load may be overwhelming
consumers’ limited resources devoted to ad processing under natural exposure conditions, but the
nonmonotonic relationship is not as clear as it was with Groupe Mu’s taxonomy.
Among taxonomies that include both visual schemes and visual tropes, Durand did not offer
predictions of specific category differences in processing. However, McQuarrie and Mick (1996)
predicted increased processing from ads without RF to schemes to tropes. This was confirmed
in laboratory experiments with verbal RFs (Mothersbaugh et al. 2002) and visual RFs (McQuarrie
and Mick 1999). It is also evident with actual ads processed under natural exposure conditions
in the current study, as Starch scores also increase along this continuum.
Additional analyses
One reviewer suggested presenting a comparative table that would show how the different
categories identified among the different taxonomies actually relate to each other. Table 4 takes
Durand’s taxonomy, because it has the greatest domain coverage, and shows the relationships
between his categories and visual RF categories from the other taxonomies. For simplicity
of presentation, only categories in which there is both some sort of conceptual overlap between
18
Table 4. Conceptual overlap and correlation of Durand categories with categories from other taxonomies.
Addition r Suppression r Substitution r Exchange r
Identity McQuarrie and Mick— McQuarrie and Mick— Group Mu—IAD .160b Gkiouzepas and Hogg—
Repetition Scheme .622b Substitution Trope .391b McQuarrie and Mick— AS-JUX .273b
Phillips and McQuarrie— Substitution Trope .610b McQuarrie and Mick—
CON-REP .175b Substitution Trope .395b
Phillips and McQuarrie—
SIM-JUX .080a
Similarity McQuarrie and Mick— Gkiouzepas and Hogg— Forceville—MP1 .232b Forceville—MP1 .069a
Repetition Scheme .792b REP-SYNTH .613b Forceville— Forceville—MP2 .206b
Gkiouzepas and Hogg- McQuarrie and Mick— Pictorial similes .125b Forceville—
b b
RS-JUX .713 Substitution Trope .149 Group Mu—IAD .064a Pictorial similes .096b
B. A. HUHMANN AND P. A. ALBINSSON
Group Mu—IPD .105b Phillips and McQuarrie— McQuarrie and Mick— Gkiouzepas and Hogg—
Phillips and McQuarrie— SIM-REP .065a Destabilization trope .516b AS-SYNTH .093b
CON-JUX .120b Phillips and McQuarrie— Group Mu—IPC .206b
SIM-JUX .268b McQuarrie and Mick—
Phillips and McQuarrie— Destabilization Tope .354b
SIM-REP .232b Phillips and McQuarrie—
SIM-FUS .196b
Difference Gkiouzepas and Hogg—
RS-SYNTH .803b
Group Mu—IAD .141b
McQuarrie and Mick—
Substitution Trope .449b
Opposition McQuarrie and Mick— Phillips and McQuarrie— Group Mu—IAD .096b Forceville—MP1 .070a
Reversal scheme .941b OPP-REP .265b Forceville—MP2 .378b
Phillips and McQuarrie— McQuarrie and Mick—
OPP-JUX .620b Destabilization trope .589b
Phillips and McQuarrie—
OPP-FUS .159b
False Phillips and McQuarrie— McQuarrie and Mick— McQuarrie and Mick— Forceville—MP2 .167b
Homology OPP-JUX .110b Substitution Trope .412b Destabilization trope .289b McQuarrie and Mick—
Destabilization trope .094b
Phillips and McQuarrie
—OPP-REP .071a
Note. Significance of correlation between visual RF categories: ap .05, bp .01, cp .001.
JOURNAL OF CURRENT ISSUES & RESEARCH IN ADVERTISING 19
the category definitions and a significant correlation coefficient appear in Table 4. Although
a few instances in which the definitions seemed to have some conceptual overlap did not exhibit
significant correlations, significant correlations were not observed without conceptual overlap.
Of course, given the different bases for developing the six taxonomies (e.g., classical rhetoric,
pictorial rhetoric, visual structures), visual RF categories do not directly map from one taxonomy
to another. For example, McQuarrie and Mick’s repetition schemes relate to two Durand
categories (ADD-ID and ADD-SIM).
Discussion
Advertising is in the unique position of a multimodal communication with both images and
language commonly available for rhetorical use, unlike media limited to the language mode, such
as speech or writing. Thus, in addition to verbal RF taxonomies (e.g., McGuire 2000; McQuarrie
and Mick 1996) that have arisen from classical rhetoric, advertising researchers have also sought
to develop taxonomies that encompass the persuasive possibilities of visual RFs in advertising
images (e.g., Durand 1970, 1987; Forceville 1996, 2007; Gkiouzepas and Hogg 2011; Groupe Mu
1992; McQuarrie and Mick 1999, 2003; Phillips and McQuarrie 2004). The current study sought to
compare these six visual RF taxonomies in terms of criteria for a useful taxonomy by examining
their parsimony, domain coverage, reliability, and predictive validity (Hunt 1991; Weber 2012).
To aid comparison across the six visual RF taxonomies, Table 5 summarizes the results of the
five research questions. This section next discusses the trade-offs for the various visual RF taxono-
mies in reference to these criteria of parsimony, domain coverage, reliability, and predictive validity.
Because they were designed to include only visual tropes, not visual schemes, the taxonomies
of Groupe Mu, Forceville, Gkiouzepas and Hogg, and Phillips and McQuarrie by design offer less
domain coverage than Durand’s or McQuarrie and Mick’s taxonomies. As a reminder, visual
schemes arrange formal image characteristics (e.g., lines, shapes, colors, objects, or symbols) to
by cloth, paint, or another substance that reveals enough of the form to indicate a person’s pres-
ence. Despite the variety of forms, reliability for this category was still exceptional,
although lower.
Also, van Mulken (2003) compared two visual RF taxonomies—those of Groupe Mu and of
McQuarrie and Mick—in a content analysis with a similar sized sample of ads. However, the cur-
rent study examines the performance of a wider set of visual RF taxonomies from the advertising
literature on more criteria, including their ability to account for differences in ad processing,
which was not addressed in van Mulken’s content analysis. Beyond the ability to reliably identify
and classify visual rhetoric, what if any effect do visual RFs have on consumer processing of ad
information? Designing effective advertisements is of paramount importance to practitioners, and
the study of techniques to improve ad effectiveness continues to interest advertising researchers.
Thus, the comparison across taxonomies of visual RFs’ ability to affect advertising readership
scores should appeal to both practitioners and researchers. This is an important contribution
because it verifies the power of visual RFs in actual ads under natural exposure conditions to
enhance ad processing. Previously, research had only observed this improvement in ad processing
outcomes in laboratory experiments. The current study measured ad processing via Starch ad
readership scores, which indicate the degree of attention, brand processing, and copy readership
that an ad receives.
All six visual RF taxonomies showed higher Starch scores overall for ads classified by that tax-
onomy as containing visual RFs than those for ads without visual RFs. However, not all visual RF
taxonomies improved Starch scores at the level of the individual categories. Upon examining ad
processing effects of each taxonomy’s individual categories, only with Forceville’s and McQuarrie
and Mick’s taxonomies did ads within every category improve at least one of the three Starch
scores over ads without visual RFs. Also, only McQuarrie and Mick’s and Groupe Mu’s taxono-
mies showed predictive validity in at least half their categories supporting theorized processing
differences.
Regardless of the taxonomy, visual RFs had the greatest effect on attention to advertising, a
moderate effect on brand association, and a weaker, though often still significant, effect on extent
of ad processing or copy readership. Visual RFs, therefore, generally encourage greater noting,
attention, and brand associations, but can also sometimes pique consumers’ curiosity in ways that
promote greater investigation of the ad message, which fits the theoretical perspective on incon-
gruity and meaning openness advanced in Huhmann (2007) and Mothersbaugh et al. (2002).
obtaining a larger sample size of copy-tested ads or a stratified sampling procedure to ensure
adequate numbers of observations within every category.
Finally, beyond the development of visual RF taxonomies, much of the extant research on vis-
ual RFs in advertising has focused on the effects of different visual metaphors (e.g., McQuarrie
and Phillips 2005; Mohanty and Ratneshwar 2016; Phillips 1997, 2000; Phillips and McQuarrie
2002). More research is needed to better understand ad processing effects of other types of visual
RFs and, in particular, visual schemes.
Conclusion
In conclusion, work on rhetoric, including that on visual RFs, has provided fruitful insights and
useful distinctions for advertising research. The current assessment further advances the state of
the visual rhetoric literature by comparing six competing visual RF taxonomies to determine
which one exhibits the best combination of domain coverage, adequate reliability, parsimony, and
predictive validity. McQuarrie and Mick’s taxonomy exhibits the best balance of Hunt’s (1991)
criteria for a useful taxonomy. It has the added advantage of applicability with both verbal and
visual modalities. In fact, Phillips and McQuarrie (2002) used it to content analyze both visual
and verbal rhetoric. This is not to say that the other taxonomies lack usefulness. For practitioners
interested in designing ad images, both Durand’s and Phillips and McQuarrie’s taxonomies have
benefits because they more precisely identify specific types of visual RFs. Unlike Hill’s (2004)
assertion that tools developed for verbal rhetoric cannot effectively be applied in a visual rhetoric
context, this study concurs with previous studies (Phillips and McQuarrie 2002; van Mulken
2003) that, at least with regard to RFs, tools such as McQuarrie and Mick’s taxonomy based on
millennia of study of classical rhetoric can be effectively applied to the visual modality. Finally,
the current study also helps answer Groupe Mu’s (1970, 27) call for a “transrhetoric” that would
go beyond classification to seek the explanation, effect, and value of different rhetorical figures
because the current study documents strong effects of visual RFs on ad attention and smaller, but
often still significant, effects on ad copy readership.
References
Calkins, Earnest. 1928. Business the Civiliser. Boston, MA: Little, Brown.
Callister, Mark A., and Lesa A. Stern. 2007. “Inspecting the Unexpected: Schema and the Processing of Visual
Deviations.” In Go Figure! New Directions in Advertising Rhetoric, edited by Edward F. McQuarrie and
Barbara J. Phillips, 137–156. Armonk, NY: M.E. Sharpe.
Durand, Jacques. 1987. “Rhetorical Figures in the Advertising Image.” In Marketing and Semiotics: New Directions
in the Study of Signs for Sale, edited by Donna Jean Umiker-Sebeok, 295–318. New York, NY: Mouton de
Gruyte.
Durand, Jacques. 1978. “Figures de rhetorique et image publicitaire: Compte rendu d’une recherche.” Humanisme
et entreprise 110:25–34.
Durand, Jacques. 1970. “Rhetoric and Advertising Image.” Communications 15 (1):70–95.
Dyer, Gillian. 1982. Advertising as Communication. London, UK: Routledge.
Finn, Adam. 1988. “Print Ad Recognition Readership Scores: An Information Processing Perspective.” Journal of
Marketing Research 25 (2):168–77.
Forceville, Charles. 2007. “Pictorial and Multimodal Metaphors in Commercials.” In Go Figure! New Directions in
Advertising Rhetoric, edited by Edward F. McQuarrie and Barbara J. Phillips, 178–204. Armonk, NY: M.E.
Sharpe.
Forceville, Charles. 1996. Pictorial Metaphor in Advertising. London, UK: Routledge
Franke, George R., Bruce A. Huhmann, and David L. Mothersbaugh. 2004. “Information Content and Consumer
Readership of Print Ads: A Comparison of Search and Experience Products.” Journal of the Academy of
Marketing Science 32 (1):20–31.
Funk, V. A. 1983. “The Value of Natural Classification.” In Numerical Taxonomy, edited by Joseph Felsenstein,
18–21. New York, NY: Springer-Verlag.
24 B. A. HUHMANN AND P. A. ALBINSSON
Gkiouzepas, Lampros, and Margaret K. Hogg. 2011. “Articulating a New Framework for Visual Metaphors in
Advertising.” Journal of Advertising 40 (1):103–20.
Gray, Stephanie A., and Rita Snyder. 1989. “Metaphor in Advertising: Effects on Memory.” In Proceedings of the
Society for Consumer Psychology, edited by Meryl P. Gardner, 85–87. Washington, DC: American
Psychological Association.
Groupe Mu. 1992. Traite du Signe Visual. Pour une Rhetorique de L’image. Paris, France: Seuil.
Groupe Mu. 1980. “Plan D’une Rhetorique De L’image.” Kodikas Code 2:249–68.
Groupe Mu. 1970. Rhetorique Generale. Paris, France: Larousse.
Halkias, Georgios, and Flora Kokkinaki. 2014. “The Degree of Ad-Brand Incongruity and the Distinction between
Schema-driven and Stimulus-driven Attitudes.” Journal of Advertising 43 (4):397–409.
Hill, Charles A. 2004. “The Psychology of Rhetorical Images” in Defining Visual Rhetorics, edited by Charles A.
Hill and Marguerite Helmers, 25–40. Mahwah, NJ: Erlbaum.
Huhmann, Bruce A. 2007. “A Model of the Cognitive and Emotional Processing of Rhetorical Works in
Advertising.” In Go Figure! New Directions in Advertising Rhetoric, edited by Edward F. McQuarrie and
Barbara J. Phillips, 133–175. Armonk, NY: M.E. Sharpe.
Huhmann, Bruce A., and Pia A. Albinsson. 2012. “Does Rhetoric Impact Advertising Effectiveness with Liking
Controlled?” European Journal of Marketing 46 (11/12):1476–500.
Huhmann, Bruce A., George R. Franke, and David L. Mothersbaugh. 2012. “Print Advertising: Executional Factors
and the RPB Grid.” Journal of Business Research 65 (6):849–954.
Hunt, Shelby D. 1991. Modern Marketing Theory: Critical Issues in the Philosophy of Marketing Science. Cincinnati,
OH: South-Western.
Kaplan, Stuart J. 1992. “A Conceptual Analysis of Form and Content in Visual Metaphors.” Communication
13:197–209.
Leigh, James H. 1994. “The Use of Figures of Speech in Print Ad Headlines.” Journal of Advertising 23 (2):17–33.
Linnaeus, Carl. 1750, published 1820. “Letter to the Marquis of Grimaldi.” In A Selection of the Correspondence of
Linnaeus, and Other Naturalistists, from the Original Manuscripts, Vol. 2, edited by James Edward Smith,
459–463. London, UK: Longman, Hurst, Rees, Orme, and Brown.
Maes, Alfons, and Joost Schilperoord. 2008. “Classifying Visual Rhetoric: Conceptual And Structural Heuristics,” in
Go Figure! New Directions in Advertising Rhetoric, edited by Edward F. McQuarrie and Barbara J. Phillips,
227–253. New York, NY: M.E. Sharpe.kaplan.
Maes, Alfons, and Joost Schilperoord. 2009. “Schemes and Tropes in Visual Communication:The Case of Object
Grouping in Advertisements.” In Discourse of Course: An Overview of Research in Discourse Studies, edited by
Jan Renkema, 67–79. Amsterdam, Netherlands: John Benjamin Publishing.
McGuire, William J. 2000. “Standing on the Shoulders of the Ancients: Consumer Research, Persuasion, and
Figurative Language.” Journal of Consumer Research 27 (1):109–14.
McQuarrie, Edward F., and David G. Mick. 2009. “A Laboratory Study of the Effect of Verbal Rhetoric versus
Repetition When Consumers Are Not Directed to Process Advertising.” International Journal of Advertising 28
(2):287–312.
McQuarrie, Edward F., and David G. Mick. 2003. “Visual and Verbal Rhetorical Figures under Directed Processing
versus Incidental Exposure to Advertising.” Journal of Consumer Research 29 (4):579–87.
McQuarrie, Edward F., and David G. Mick. 1999. “Visual Rhetoric in Advertising: Text-Interpretive, Experimental,
and Reader-Response Analyses.” Journal of Consumer Research 26 (1):37–54.
McQuarrie, Edward F., and David G. Mick. 1996. “Figures of Rhetoric in Advertising Language.” Journal of
Consumer Research 22 (4):424–37.
McQuarrie, Edward F., and Barbara J. Phillips. 2008. “It’s Not Your Father’s Magazine Ad: Magnitude and
Direction of Recent Changes in Advertising Style.” Journal of Advertising 37 (3):95–106.
McQuarrie, Edward F., and Barbara J. Phillips. 2005. “Indirect Persuasion in Advertising—How Consumers
Process Metaphors Presented in Pictures and Words.” Journal of Advertising 34 (2):7–20.
Messaris, Paul. 1997. Visual Persuasion: The Role of Images in Advertising. Thousand Oaks, CA: Sage.
Messaris, Paul. 2009. “What’s Visual about ‘Visual Rhetoric’?” Quarterly Journal of Speech 95 (2):210–23.
Mohanty, Praggyan (Pam), and S. Ratneshwar. 2016. “Visual Metaphors in Ads: The Inverted-U Effects of
Incongruity on Processing Pleasure and Ad Effectiveness.” Journal of Promotion Management 22 (3):443–60.
Mothersbaugh, David L., Bruce A. Huhmann, and George R. Franke. 2002. “Combinatory and Separative Effects of
Rhetorical Figures on Consumers’ Effort and Focus in Ad Processing.” Journal of Consumer Research 28
(4):589–602.
Nunnally, Jum C. 1978. Psychometric Theory. 2nd ed. New York, NY: McGraw-Hill.
Perreault, William D., Jr., and Laurence E. Leigh. 1989. “Reliability of Nominal Data Based on Qualitative
Judgments.” Journal of Marketing Research 26 (2):135–48.
Phillips, Barbara J. 2003. “Understanding Visual Metaphor in Advertising.” In Persuasive Imagery: A Consumer
Perspective, edited by Linda M. Scott and Rajeev Batra, 297–310. Mahwah, NJ: Erlbaum.
JOURNAL OF CURRENT ISSUES & RESEARCH IN ADVERTISING 25
Phillips, Barbara J. 2000. “The Impact of Verbal Anchoring on Consumer Response to Image Ads.” Journal of
Advertising 29 (1):15–24.
Phillips, Barbara J. 1997. “Thinking into It: Consumer Interpretation of Complex Advertising Images.” Journal of
Advertising 26 (2):77–87.
Phillips, Barbara J., and Edward F. McQuarrie. 2004. “Beyond Visual Metaphor: A New Typology of Visual
Rhetoric in Advertising.” Marketing Theory 4 (1–2):113–36.
Phillips, Barbara J., and Edward F. McQuarrie. 2002. “The Development, Change, and Transformation of
Rhetorical Style in Magazine Advertisements, 1954–1999.” Journal of Advertising 31 (4):1–13.
Rust, Roland T., and Bruce Cooil. 1994. “Reliability Measures for Qualitative Data: Theory and Implications.”
Journal of Marketing Research 31 (1):1–14.
Scott, Linda M. 1994. “Images in Advertising: The Need for a Theory of Visual Rhetoric.” Journal of Consumer
Research 21 (2):252–73.
Theodorakis, Ioannis G., Christos Koritos, and Vlasis Stathakopoulos. 2015. “Rhetorical Maneuvers in a
Controversial Tide: Assessing the Boundaries of Advertising Rhetoric.” Journal of Advertising 44 (1):14–24.
Tom, Gail, and Anmarie Eves. 1999. “The Use of Rhetorical Devices in Advertising.” Journal of Advertising
Research 39:39–43.
Toncar, Mark, and James Munch. 2001. “Consumer Responses to Tropes in Print Advertising.” Journal of
Advertising 30 (1):55–65.
van Mulken, Margot. 2003. “Analyzing Rhetorical Devices in Print Advertisements.” Document Design 4
(2):114–28.
van Mulken, Margot, Andreu van Hooft, and Ulrike Nederstigt. 2014. “Finding the Tipping Point: Visual
Metaphor and Conceptual Complexity in Advertising.” Journal of Advertising 43 (4):333–43.
Weber, Rob. 2012. “Evaluating and Developing Theories in the Information Systems Discipline.” Journal of the
Association for Information Systems 13 (1):1–30.