The Associative-Propositional Evaluation Model: Operating Principles and Operating Conditions of Evaluation
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
THE APE MODEL 1
The Associative-Propositional Evaluation Model:
Operating Principles and Operating Conditions of Evaluation
Bertram Gawronski Galen V. Bodenhausen
The University of Western Ontario Northwestern University
Introduction the behavioral outcome of propositional processes.
Conflicting evaluative responses to the same By making specific assumptions about mutual
object, individual, or social group can be vexingly interactions between associative and propositional
commonplace in everyday life (Conner & Armitage, processes, the APE model implies a wide range of
2008). For example, we may experience spontaneous predictions about the conditions under which implicit
apprehension and discomfort when encountering and explicit evaluations show either converging or
members of stigmatized groups even though we diverging patterns of responses. These predictions
intellectually abhor prejudice and wish to express and relevant empirical evidence are extensively
solidarity with minorities. Analogously, people can reviewed elsewhere (Gawronski & Bodenhausen,
feel strong romantic attraction toward another person 2011), and we therefore refrain from providing
despite firmly believing that this person is not a good elaborate discussions of these aspects in the current
match. Although these two examples are quite chapter. Instead, we focus on how the distinction
distinct, both are characterized by a conflict between between associative and propositional processes
spontaneous evaluative responses and deliberate relates to the distinction between automatic and
evaluative judgments. One valuable approach to controlled processes that has shaped dual-process
studying such evaluative conflicts is to compare theorizing for the last three decades (see Gawronski
people’s responses on traditional self-report measures & Creighton, in press). Our central argument is that
(e.g., attitude scales, likeability ratings) to their the associative-propositional distinction refers to the
responses on performance-based paradigms (e.g., principles that define what a particular process is
implicit association test, sequential priming). doing. In contrast, the automatic-controlled
Conceptually, deliberate evaluative judgments on the distinction refers to empirical claims about when that
former type of measures can be described as explicit process is operating. Although the two dichotomies
evaluations in the sense that their evaluative meaning are sometimes assumed to overlap, the APE model
is explicit in the observed response (e.g., participants draws a sharp line between operating principles and
explicitly report their agreement or disagreement operating conditions (Gawronski & Bodenhausen,
with an evaluative statement about an attitude 2007, 2009). In this spirit, we first explain the
object). Conversely, spontaneous evaluative defining features of associative and propositional
responses on the latter type of measures can be processes and then outline the manner in which the
described as implicit evaluations in the sense that two processes are assumed to operate in an automatic
their evaluative meaning is implicit in the observed versus controlled fashion. To resolve some common
response (e.g., evaluative responses are inferred from misconceptions about the APE model, we also
participants’ latencies in responding to positive and address the difference between our dual-process
negative words that are preceded by brief distinction in terms of associative and propositional
presentations of an attitude object). Over the past processes and cognate conceptualizations in terms of
years, a substantial body of research has documented dual-systems and dual-representations.
the possibility of dissociations between implicit and Operating Principles
explicit evaluations, including different antecedents, As we noted above, the associative-propositional
different consequences, and discrepant evaluations of distinction refers to the operating principles that
the same object. define what a given process is doing. In the APE
The main goal of the current chapter is to review model, we specify associative processes as the
the core assumptions of our associative-propositional activation of mental associations on the basis of
evaluation (APE) model, which explains feature similarity and spatio-temporal contiguity;
dissociations between implicit and explicit propositional processes are defined as the validation
evaluations in terms of their underlying mental of momentarily activated information on the basis of
processes (Gawronski & Bodenhausen, 2006a, logical consistency. These general definitions can be
2006b, 2007, 2011). Whereas implicit evaluations are further specified on the basis of whether they refer to
assumed to be the behavioral outcome of associative the expression or the formation of evaluative
processes, explicit evaluations are conceptualized as representations.
To appear in: J. W. Sherman, B. Gawronski, & Y. Trope (Eds.). Dual-process theories of the social mind. New York: Guilford Press.THE APE MODEL 2
Expression of Evaluative Representations However, the activation of associated concepts is not
A central assumption in research on attitudes is entirely context-driven, because it is constrained by
that encountering a stimulus can elicit a positive or the preexisting structure of mental links in memory.
negative response by activating stored evaluative After all, different contexts can modulate the
representations in memory. In the APE model, we activation of concepts in response to a given object
assume that principles of similarity matching only if these concepts are part of the mental
determine the activation of mental concepts that representation of that object.
represent the encountered stimulus (e.g., Afrocentric A central feature of associative activation is that it
features of a face activating the concept African is independent of subjective truth or falsity.
American), which can spread to other concepts that Specifically, we assume that the principles of
are associatively linked with the stimulus (e.g., similarity matching determine the activation of
activation of the concept African American spreading associated concepts regardless of whether the
to the associated stereotypical attribute hostile). To activated link is considered valid or invalid. For
the extent that the associated concepts have a positive example, encountering a Muslim-looking man may
or negative connotation, their activation is assumed activate the concept terrorism even if a person rejects
to produce a spontaneous gut response that is in line the implied connection between Muslims and
with the valence of these concepts (implicit terrorism (Devine, 1989). According to the APE
evaluation). model, the validity of activated links is determined by
An important aspect of the similarity matching a process of propositional validation. Specifically, we
principle is that stimuli do not have to be perceptually assume that activated information is regarded as valid
identical across time and contexts to elicit the same unless the default process of affirming the validity of
evaluative response. Instead, configurations of input activated information produces an inconsistent set of
stimuli that pass a critical threshold of similarity are beliefs. The central idea underlying these
sufficient to activate the same mental representation assumptions is that, although consistency does not
(Smith, 1996). For example, relatively Afrocentric guarantee accuracy, inconsistency is an unambiguous
facial features can activate Black stereotypes even indicator of an erroneous component in one’s system
when they are present in the faces of individuals who of beliefs (Gawronski, in press). In such cases, the
are categorized as White (Blair, Judd, & Fallman, momentarily considered set of information needs to
2004). In addition, the principle of similarity be updated, which involves a reassessment of the
matching implies that even unknown stimuli can validity of each component.
elicit spontaneous evaluative responses to the extent With regard to evaluative responses, we assume
that they resemble a previously encountered stimulus that the affective gut reactions resulting from
with a stored evaluative representation. For example, associatively activated concepts are translated into
unknown individuals may elicit a spontaneous the format of a propositional statement (e.g., a
positive or negative response by virtue of their negative affective reaction toward object X is
resemblance to people we know (Gawronski & transformed into propositional statements such as “I
Quinn, in press). dislike X” or “X is bad”). To the extent that the
Another important aspect of the similarity propositional evaluation implied by an affective gut
matching principle is that associative activation is not response is consistent with other momentarily
an all-or-none process, such that encountering a considered propositions, it may be endorsed in a
given object would activate each and every concept verbal evaluative judgment (explicit evaluation). If,
that is associated with that object in memory. Instead, however, the overall set of momentarily considered
objects tend to activate only a limited subset of propositions is inconsistent, the inconsistency has to
associated concepts. Which subset of associated be resolved to avoid aversive feelings of dissonance
concepts is activated in response to a given object is (Festinger, 1957). In general, propositional
assumed to depend on the overall configuration of evaluations of a given object may be assessed for
input stimuli, including both the target object and the their consistency with (a) non-evaluative propositions
context in which it is encountered. For example, about states of affairs and (b) propositional
encountering a Black person in a jazz bar may evaluations of other attitude objects (Jones & Gerard,
activate the stereotypical attribute musical, whereas 1967). To the extent that a set of momentarily
the same Black person may activate the stereotypical considered propositions is inconsistent, consistency
attribute criminal if that person is encountered in a may be restored either by rejecting one of the
dark alley (for a review, see Gawronski & Sritharan, involved propositions (i.e., reversing the subjective
2010). Hence, an attitude object may elicit distinctly truth value of that proposition) or by searching for an
different implicit evaluations depending on the additional proposition that resolves the inconsistency
particular context in which it is encountered. (Gawronski & Strack, 2004). For example, the
To appear in: J. W. Sherman, B. Gawronski, & Y. Trope (Eds.). Dual-process theories of the social mind. New York: Guilford Press.THE APE MODEL 3
propositional implication of a negative affective involve an affirmation of new information. For
reaction to minority members (e.g., “I dislike African example, affirming the proposition “old people are
Americans.”) may be inconsistent with the good drivers” may strengthen the association
propositional evaluation of another attitude object between the concepts old people and good drivers,
(e.g., “Negative evaluations of disadvantaged groups thereby increasing the correspondence between
are wrong.”) and non-evaluative propositions about implicit and explicit evaluations. Thus, if a person is
states of affairs (e.g., “African Americans are a motivated to hold a positive (negative) impression of
disadvantaged group.”). Thus, the inconsistency an attitude object, but experiences a negative
between the three propositions may lead to a (positive) affective reaction toward that object, the
rejection of the negative affective reaction as a valid individual may engage in a directed memory search
basis for an evaluative judgment (e.g., “I like African to retrieve positive (negative) information about the
Americans.”). However, consistency may also be object, which should promote a positive (negative)
restored by rejecting either the non-evaluative evaluation for both explicit and implicit evaluations.
proposition about states of affairs (e.g., “African If, however, the positive (negative) impression is
Americans are not a disadvantaged group.”) or the maintained by merely negating the negative
propositional evaluation of another relevant attitude (positive) evaluation implied by the affective gut
objects (e.g., “Negative evaluations of disadvantaged response (i.e., without retrieving supportive positive
groups are okay.”). Whereas the former case should or negative information), explicit and implicit
result in a dissociation between implicit and explicit evaluations should show a dissociation, such that
evaluations, the two kinds of evaluations should explicit evaluations reflect the desired positive
show converging negative responses in the latter (negative) evaluation, whereas implicit evaluations
cases (Gawronski, Peters, Brochu, & Strack, 2008). should reflect the original negative (positive)
In addition to such “bottom-up” effects of response.
associative on propositional processes, the APE Formation of Evaluative Representations
model also includes specific assumptions about “top- Before an evaluative representation can be
down” effects of propositional thinking on activated, it has to be formed on the basis of some
associative processes. Specifically, we assume that kind of learning experience. In the APE model, we
processes of propositional reasoning can influence distinguish between two conceptually distinct
associative processes by activating new information processes of forming evaluative representations
in the course of validating activated information. For depending on whether they are based on associative
example, if people are motivated to believe in the or propositional principles. Drawing on our general
validity of a particular proposition, they may engage definition of associative processes, associative
in a selective search for information that supports the learning can be specified as the formation of
validity of that proposition (Kunda, 1990). In such associative links between mental concepts on the
cases, biased retrieval of information from memory basis of observed spatio-temporal contiguities. The
can activate associated concepts of a particular central assumption underlying this definition is that
valence, which produces correspondence between observed co-occurrences between objects and events
implicit and explicit evaluations in a “top-down” result in a co-activation of their corresponding mental
fashion (Peters & Gawronski, 2011a). concepts, which in turn creates an associative link
An important factor in such top-down effects is between the two. Repeatedly observing the same co-
whether propositional reasoning involves an occurrences strengthens this link, which facilitates
affirmation or negation of the relevant information. the spread of activation from one concept to the other
Specifically, we argue that merely negating a upon encountering one of the two associated stimuli.
particular proposition (i.e., reversing its truth value) An illustrative example of associative learning is
is insufficient to deactivate the associative link evaluative conditioning (EC; see De Houwer,
underlying this proposition. In fact, negations often Thomas, & Baeyens, 2001), in which repeated
lead to ironic effects, such that the activation level of pairings of a conditioned stimulus (CS) with a
the underlying association is enhanced rather than positive or negative unconditioned stimulus (US) can
reduced (e.g., Gawronski, Deutsch, Mbirkou, Seibt & produce a mental association between the CS and the
Strack, 2008). For example, negating the proposition US in memory. As a result, subsequent presentations
“old people are bad drivers” may enhance the of the CS spread to the representation of the US,
associative link between the concepts old people and which produces an evaluative response to the CS that
bad drivers, thereby leading to a dissociation is in line with the valence of the US (e.g., Walther,
between implicit and explicit evaluations (Deutsch, Gawronski, Blank, & Langer, 2009).
Gawronski, & Strack, 2006). This situation is In contrast to the associative principle of mere co-
different if processes of propositional reasoning activation, propositional learning is defined as the
To appear in: J. W. Sherman, B. Gawronski, & Y. Trope (Eds.). Dual-process theories of the social mind. New York: Guilford Press.THE APE MODEL 4
formation of evaluative representations on the basis An important aspect of the distinction between
of propositional information that is regarded as valid. associative and propositional learning is that the same
This definition is based on our conceptualization of stimulus event may influence evaluative
propositional processes as being concerned with the representations through two simultaneously operating
validity of momentarily activated information. processes (Gawronski & Bodenhausen, 2006a). For
Propositional learning may be based on new example, repeated co-occurrences of a CS and a
information that is presented in the format of valenced US may create a mental link between the
propositional statements (e.g., persuasive arguments two stimuli through processes of associative learning.
asserting the quality of a consumer product). At the same time, the observed co-occurrences may
Alternatively, propositional learning can be based on provide the basis for self-generated propositions
propositional inferences about observed stimulus about their evaluative meaning (e.g., propositional
events in the environment (e.g., co-occurrences inferences about the CS being a cause of the positive
between stimulus events can trigger propositional or negative event represented by the US), which may
inferences about their causal relation). Whereas the influence the evaluative representation through
former case involves the acquisition of externally processes of propositional learning. This distinction
provided propositional information, the latter case is important, because it qualifies the conceptual
involves the acquisition of self-generated equation of evaluative conditioning and associative
propositional information. Yet, in both cases, the new learning that we endorsed in the initial presentation
information has to pass a process of propositional of the APE model (for a more detailed discussion, see
validation. This validity assessment is equivalent to Gawronski & Bodenhausen, 2011). To the extent that
the one involved in the expression of evaluative EC can be defined as the change in the evaluation of
representations, such that new propositional a CS due to its pairing with a valenced US (De
information may be regarded as either true or false Houwer, 2007), EC effects may be the result of either
depending on its consistency with other momentarily associative or propositional learning (e.g.,
considered propositions. Gawronski, Balas, & Creighton, 2012). Yet, a
Although associative and propositional learning theoretical challenge is to identify the conditions
represent distinct mechanisms of forming evaluative under which the effect of observed co-occurrences on
representations, their outcomes are assumed to evaluative responses is mediated by associative or
interact in a manner that is similar to the mutual propositional learning (or both). This question
interactions in the expression of evaluative pertains to the operating conditions of associative and
representations. First, associatively formed propositional processes, which we discuss in the
representations may provide the input for following section (cf. Gawronski & Bodenhausen,
propositional inferences, implying a “bottom-up” 2007, 2011).
effect of associative on propositional processes. Operating Conditions
Thus, whether or not the evaluation implied by an Whereas the associative-propositional distinction
associatively formed representation is regarded as refers to the operating principles that define what a
valid depends on the consistency of this evaluation particular process is doing, the automatic-controlled
with other momentarily considered propositions (e.g., distinction refers to empirical claims about when that
Gawronski & LeBel, 2008). Second, propositional process is operating (e.g., when there is no conscious
processes may influence associative processes in a awareness; when there is no intention to start the
“top-down” fashion when externally provided or self- process; when cognitive resources are reduced; when
generated propositions create new mental links in there is a goal to alter or stop the process; see Bargh,
memory (e.g., Whitfield & Jordan, 2009). As we 1994). According to the APE model, there is no one-
outlined above, an important determinant of such top- to-one mapping between operating principles and
down effects is whether the involved inferences entail operating conditions, such that associative processes
an affirmation or negation of the relevant would operate automatically, whereas propositional
information. Whereas the affirmation of a given processes operate in a controlled fashion (Gawronski
proposition should create an evaluative representation & Bodenhausen, 2009). Instead, both associative and
that is in line with the meaning of that proposition, propositional processes have automatic and
negating a given proposition is claimed to have ironic controlled aspects. Moreover, each type of process
effects. 1 involves different components, which require
1
Challenging the generality of such ironic effects, a recent study to be systematic limits in the processing of observed co-
by Peters and Gawronski (2011b) found that negation can qualify occurrences between stimuli that involve a contrastive relation
the stored representations resulting from subjectively invalid (e.g., a stimulus preventing a negative stimulus becoming
propositions to the extent that the invalidation occurs within a associated with the negative stimulus; see Moran & Bar-Anan, in
sufficiently short interval after encoding. Nevertheless, there seem press).
To appear in: J. W. Sherman, B. Gawronski, & Y. Trope (Eds.). Dual-process theories of the social mind. New York: Guilford Press.THE APE MODEL 5
separate consideration in a thorough analysis of unaware of the particular associations that are
operating conditions. Because different features of responsible for these reactions.
automatic processing need not co-occur, we also As for propositional processes, we assume that
deem it important to distinguish between the unique conscious awareness is not required for the default
roles of awareness, intentionality, efficiency, and process of affirming the validity of activated
controllability (see Table 1). Importantly, whereas information, although people may sometimes engage
the distinction between associative and propositional a conscious reassessment of the validity of that
processes is purely conceptual, any claims about their information. Similar considerations apply to the
operating conditions are empirical and therefore have process of monitoring the consistency of
to be assessed on the basis of relevant evidence. momentarily activated information. In many cases,
Awareness this monitoring process may operate outside of
Expression. Conscious awareness is commonly conscious awareness, even though people can
defined in terms of introspective access to mental certainly monitor their belief systems consciously to
processes or mental contents. Empirically, lack of identify potential inconsistencies. However,
introspective access can be established through inconsistency between activated information is
participants’ inability to verbally report a mental assumed to raise conscious awareness, which in turn
process or mental content. In the APE model, we supports the resolution of inconsistency (Morsella,
argue that people usually have experiential access to Zarolia, & Gazzaley, 2012). In such cases, the
their affective gut reactions resulting from necessary reassessment of the activated information
associatively activated concepts, and that they often involves conscious awareness of the involved
rely upon these reactions in making propositional processing steps, such as the negation (i.e., reversal
evaluative judgments. Still, people also sometimes of the truth value) of a particular proposition or the
reject their affective gut reactions as a basis for an search for information that resolves the
evaluative judgment when these reactions are inconsistency. The behavioral process of reporting an
inconsistent with other momentarily considered evaluative judgment generally occurs under
propositions. However, such dissociations between conscious awareness.
affective gut reactions and evaluative judgments do Formation. Associative learning is commonly
not imply that the affective gut reactions are assumed to be independent of people’s awareness of
introspectively inaccessible. the relevant contiguities that are responsible for the
Note, however, that although people may be formation of new associative links. The APE model
experientially aware of the affective gut reactions generally agrees with this contention. Yet, several
resulting from activated associations, they may studies found that EC effects were smaller (e.g.,
sometimes be unaware of the processes that gave rise Hofmann, De Houwer, Perugini, Baeyens, &
to these reactions (Gawronski, Hofmann, & Wilbur, Crombez, 2010) or fully attenuated (e.g., Pleyers,
2006). For example, people may show a positive or Corneille, Luminet, & Yzerbyt, 2007) when
negative gut response toward an unfamiliar participants failed to remember the relevant CS-US
individual on the basis of that person’s similarity to a pairings. These findings have led some researchers to
known individual. However, they may not be able to reject the hypothesis that observed CS-US
identify the similarity between the two individuals as contiguities can influence evaluative representations
the cause of their affective gut response (e.g., through an unconscious process of associative
Günyadin, Zayas, Selcuk, & Hazan, 2012). In terms learning (e.g., Mitchell, De Houwer, & Lovibond,
of the APE model, such effects can be explained by 2009). However, there are a number of
the principle of similarity matching that characterizes methodological issues that have to be taken into
the process of associative activation. In the current account when interpreting the relation between
example, the resemblance between the two memory performance and evaluation. First, it is
individuals may activate evaluative concepts that are important to note that current approaches to measure
associated with the known individual, thereby memory for CS-US pairings confound effects of
eliciting an affective gut response that is in line with contingency memory and conditioned attitudes,
the valence of these concepts. Yet, the particular thereby biasing results in favor of the conclusion that
content of these associations may remain contingency memory is a necessary precondition for
unconscious even when people are experientially EC effects. If the confounded influences are
aware of the affective gut reaction resulting from disentangled by means of process dissociation
these associations. In other words, people may be (Jacoby, 1991), EC effects tend to emerge even in
experientially aware of their affective gut reactions to absence of contingency memory (Hütter, Sweldens,
a person or object, but they may sometimes be Stahl, Unkelbach, & Klauer, 2012). Second, although
memory measures are certainly suitable to investigate
To appear in: J. W. Sherman, B. Gawronski, & Y. Trope (Eds.). Dual-process theories of the social mind. New York: Guilford Press.THE APE MODEL 6
the contribution of propositional knowledge of CS- mechanism that can produce representational
US relations to EC effects at the time of expressing changes.
an evaluative response, they remain ambiguous about Intentionality
the role of conscious awareness during the formation Expression. In general terms, a process can be
of an evaluative representation (Gawronski & described as unintentional if it is instigated in the
Walther, 2012). This ambiguity is due to the fact that absence of a person’s intention to start that process
(a) memory measures are unable to distinguish (Bargh, 1994). Applied to evaluative responses, it has
between encoding-related and retrieval-related been argued that the activation of evaluative
effects, and (b) the relation between memory associations—and thus the affective gut reactions
performance and evaluation is merely correlational, resulting from these associations—occurs regardless
which limits conclusions about causal relations of the intention to evaluate an object (e.g., Bargh,
between memory performance and EC effects. As a Chaiken, Raymond, & Hymes, 1996). In line with
result, any possible data pattern can be interpreted in this contention, the APE model assumes that the
at least two different ways, which undermines the activation of evaluative associations in memory can
suitability of memory measures to investigate indeed occur unintentionally, thereby meeting the
whether evaluative learning can occur outside of second criterion of automaticity. However, it is
conscious awareness. As outlined in detail by important to note that evaluative associations can also
Gawronski and Walther (2012), a stringent test of this be activated intentionally. For example, an individual
hypothesis requires experimental approaches in may intentionally search for particular information in
which awareness of CS-US pairings is manipulated memory, and the evaluative associations that are
during encoding. 2 To be sure, memory measures can activated through this search may influence this
be very useful to study the contribution of person’s affective responses to the relevant target
propositional knowledge to EC effects at the time of object (e.g., Peters & Gawronski, 2011a). Thus, even
expressing an evaluative response (e.g., Balas & though the activation of evaluative associations—and
Gawronski, 2012). However, such influences should thus the affective gut reactions resulting from these
not be confused with unconscious influences of CS- associations—does not require intention, evaluative
US pairings at the time of forming an evaluative associations can also be activated intentionally.
representation. As for propositional processes, we argue that
Whereas associative learning is assumed to be intention is not required for the default process of
independent of conscious awareness, propositional affirming the validity of activated information, even
learning generally requires conscious awareness. This though individuals may sometimes engage in an
hypothesis is based on evidence that externally intentional reassessment of the validity of that
provided propositional statements cannot be encoded information. Similar to our assumptions about
outside of awareness (Baars, 2002; Greenwald, awareness, we assume that such intentional
1992). What is usually extracted from a propositional reassessments will occur when the overall set of
statement under suboptimal processing conditions is activated information is inconsistent. In such cases,
the presence of individual stimuli and their co- people will intentionally restore consistency by either
occurrence, but not the propositional meaning of the negating (i.e., reversing the truth value of) a
statement. Similarly, conscious awareness is assumed particular proposition or by searching for information
to be required for the self-generation of propositions that resolves the inconsistency. In addition, we argue
in response to stimulus events. Of course, to the that the monitoring of (in)consistency occurs
extent that people are consciously aware of the unintentionally (Morsella et al., 2012), even though
contiguity of two stimuli, associative and people may sometimes engage in an intentional
propositional learning may jointly influence the assessment of the consistency of their beliefs. The
evaluative representation of these stimuli (e.g., behavioral process of reporting an evaluative
Gawronski et al., 2012). Yet, lack of conscious judgment is generally intentional.
awareness should eliminate the effects propositional Formation. Associative learning can be described
learning, leaving associative learning as the only as unintentional in the sense that the learning process
itself does not require the goal to form a new
association. However, associative learning can
2
In this context, it is important to distinguish between attention certainly have intentional antecedents, such that
and awareness as two conceptually distinct aspects of encoding
CS-US relations. Although attention to the relevant stimuli is likely
people may intentionally expose themselves to
required for both associative and propositional learning, repeated co-occurrences to facilitate their acquisition
associative learning may occur in the absence of conscious (e.g., intentional exposure to pairs of words in the
awareness to the extent that the relevant CS-US pairings are in the learning of a foreign language).
focus of attention (e.g., Custers & Aarts, 2011; Field & Moore,
2005).
To appear in: J. W. Sherman, B. Gawronski, & Y. Trope (Eds.). Dual-process theories of the social mind. New York: Guilford Press.THE APE MODEL 7
Similar considerations apply to propositional extent that the amount of relevant information is
learning. Once an externally provided or self- rather small and the complexity of the required
generated proposition is considered, its content may inferences is low, the identification and resolution of
be stored in memory even when people do not have inconsistency will require few cognitive resources
the goal to memorize it. Of course, the goal to (e.g., Richter, Schroeder, & Wöhrmann, 2009). If,
memorize the content of a given proposition may however, the amount of relevant information is large
facilitate its storage, but such a goal is not a or the required inferences are relatively complex, the
necessary precondition. In fact, a substantial body of cognitive resources that are required for the
evidence suggests that alternative processing goals identification and resolution of inconsistency will be
(e.g., impression formation) can be more effective in more substantial (e.g., Martinie, Olive, & Milland,
producing a strong memory trace than memorization 2010). In other words, we do not assume that
goals (e.g., Hamilton, Katz, & Leirer, 1980). The propositional processes are cognitively effortful per
same is true for the process of self-generating a new se. Rather, processes of propositional reasoning can
proposition, which also does not require an intention be more or less effortful depending on the complexity
to instigate this process. Although self-generated of the inferences that are involved. For example, high
propositions can certainly be the result of intentional levels of complexity are often involved when
inferences about stimulus relations, they can be motivational concerns bias the direction of
purely “stimulus-driven” in the sense that they spring propositional reasoning, in that people engage in an
from activated associations when people elaborate search for information that validates a
unintentionally notice a systematic relation between desired conclusion or invalidates an undesired
stimuli. conclusion (e.g., Moreno & Bodenhausen, 1999). The
Efficiency behavioral process of reporting the outcome of these
Expression. A process can be described as validation processes usually requires few cognitive
efficient if it operates even under conditions of resources. What does require more cognitive
reduced cognitive resources. Applied to evaluative resources is the mental process of reaching an
responses, it is commonly assumed that evaluative evaluative conclusion, not the behavioral process of
associations are activated in response to a particular reporting that conclusion.
stimulus even when cognitive resources are scarce. Formation. According to the APE model, the
The APE model generally agrees with the contention formation of mental links through associative
that associative processes are highly efficient. learning is resource-independent, although attentional
However, this efficiency does not imply that distraction may sometimes disrupt associative
evaluative associations cannot be activated in an learning if it undermines the encoding of the relevant
effortful manner. Even though evaluative contiguities (e.g., Pleyers, Corneille, Yzerbyt, &
associations—and thus the affective gut reactions Luminet, 2009). Thus, when investigating the
resulting from these associations—do not require proposed resource-independence of associative
cognitive effort to become activated, they can also be learning, we deem it important to distinguish between
activated through the effortful retrieval of evaluative different aspects of working memory capacity (see
information from memory (e.g., Peters & Gawronski, Baddeley, 2010). Whereas capacity constraints on
2011a). episodic memory (e.g., concurrent rehearsal of a
As for propositional processes, we argue that the complex digit-string) should leave associative
default process of affirming the validity of activated learning unaffected, capacity constraints on
information is highly efficient in the sense that it perceptual processing (e.g., concurrent attention to
occurs even under conditions of reduced cognitive numbers in a two-back task) may reduce the effects
resources. However, the situation is different for the of associative learning to the extent that it
monitoring and the resolution of inconsistency. In undermines the perceptual encoding of the relevant
many situations, the monitoring of momentarily stimulus contiguities (see also Custers & Aarts, 2011;
activated information may occur efficiently through Field & Moore, 2005).
the unconscious and unintentional operation of the As for propositional learning, we assume that the
brain’s conflict monitoring system (Botvinic, Cohen, mere consideration of a given proposition does not
& Carter, 2004). However, when people engage in a require substantial amounts of cognitive resources,
conscious and intentional assessment of particular although greater elaboration can certainly strengthen
pieces of information, limits in working memory the resulting mental links (Craik & Lockhart, 1972).
capacity will constrain (a) how much information Nevertheless, comprehension of externally provided
they can hold simultaneously in memory and (b) the propositional information may require more
complexity of syllogistic inferences they can perform resources if this information is highly complex. The
to assess the consistency of this information. To the same is true for the self-generation of propositional
To appear in: J. W. Sherman, B. Gawronski, & Y. Trope (Eds.). Dual-process theories of the social mind. New York: Guilford Press.THE APE MODEL 8
information, given that limited cognitive resources the goal of not forming an association between the
can constrain the complexity of self-generated relevant stimuli (e.g., Gawronski et al., 2012). A
information. more complex question is whether associative
Controllability learning can be intentionally altered (rather than
Expression. Another important characteristic that stopped). Several studies have shown that EC effects
has been used to describe associative processes is the of repeated CS-US pairings can be reversed when the
notion of controllability. In technical terms, a process CSs were assumed to have a “negative” relation to
can be described as uncontrollable if it operates the USs (e.g., the CS dislikes the US; the CS is an
despite a person’s intention to terminate that process. enemy of the US; see Fiedler & Unkelbach, 2011;
Thus, whereas the intentionality criterion refers to the Förderer & Unkelbach, 2012). Although these
goal of starting a process, the controllability criterion findings seem to suggest that associative learning can
refers to the goal of altering or stopping a process be altered by higher-order processing goals, there is
(Bargh, 1994). Applied to the current question, one evidence that reversed EC effects resulting from
could argue that the activation of evaluative “negative” CS-US relations may conceal unqualified
associations—and thus of the affective gut reactions associative effects that can be uncovered under
resulting from these association—is uncontrollable if particular conditions. For example, Langer, Walther,
this activation process cannot be altered or Gawronski, and Blank (2009) presented participants
terminated. We argue that the activation of evaluative with several CS-US pairs and additional information
associations is controllable to some extent. However, indicating that the two stimuli have either a positive
the overall success in controlling the activation of relation (i.e., like each other) or a negative relation
evaluative associations is assumed to depend on the (i.e., dislike each other). Subsequently, the original
nature of the adopted control strategy. According to valence of the USs was reversed, such that
the APE model, the most critical factor in this regard participants received negative information about
is whether the adopted control strategy implies a positive USs and positive information about negative
negation of information that is already activated or an USs (see Walther et al., 2009). Participants in a
affirmation of new information. As outlined above, control condition were presented with neutral
negating the validity of an affective gut reaction is information about the USs. Results in the control
assumed to reduce the influence of this reaction on condition showed a standard EC effect when the CSs
evaluative judgments. However, it may not eliminate and the USs had a positive relation, but a reversed EC
the affective gut reaction per se, as mere negations do effect when the CSs and the USs had a negative
not necessarily deactivate the associations that gave relation (see also Fiedler & Unkelbach, 2011;
rise to this reaction. In contrast, affirming new Förderer & Unkelbach, 2012). More importantly,
evaluative information typically activates new when the original valence of the USs was reversed,
associations in memory, which tend to influence the CS evaluations failed to produce the mirror image of
affective gut reactions resulting from activated the cross-over interaction obtained in the control
associations in the intended direction. condition. Instead, there was only an unqualified
As for propositional processes, we argue that the main effect of US valence, such that CS evaluations
default process of affirming the validity of activated directly reflected the new valence of the US that had
information is generally controllable, because been paired with a given CS regardless of whether
activated information can always be negated. the relation between the two stimuli was positive or
However, the process of monitoring the negative. These results suggest that propositional
(in)consistency of activated information is most inferences regarding “negative” CS-US relations can
likely uncontrollable, in that it cannot be altered or conceal associative effects to the extent that
stopped. The process of inconsistency resolution is information about these relations is available during
generally controllable, given that people may change the encoding of CS-US pairings (for related findings,
the preferred strategy to resolve inconsistency. For see Gawronski, Walther, & Blank, 2005). However,
example, instead of rejecting one of the involved the same pairings simultaneously create an
propositions as false, a person may search for unqualified associative link between the CS and the
information that resolves the inconsistency (or the US, which can be uncovered through subsequent
other way round). The behavioral process of revaluation of the USs.
reporting an evaluative judgment is generally As for the controllability of propositional
controllable, given that a person can always report a learning, it is certainly possible to intentionally
different judgment voluntarily. invalidate an externally provided or self-generated
Formation. In the APE model, we assume that proposition. Such goal-dependent invalidation is
associative learning is uncontrollable in the sense that conceptually equivalent to the effects of motivated
observed contiguities can create mental links despite reasoning, in which people may have a desire to
To appear in: J. W. Sherman, B. Gawronski, & Y. Trope (Eds.). Dual-process theories of the social mind. New York: Guilford Press.THE APE MODEL 9
confirm or disconfirm the validity of a given vs. propositional) and operating conditions
proposition. As with negation effects in motivated (automatic vs. controlled). We also do not assume
reasoning, however, the effectiveness of intentional systematic overlap of either distinction with other
invalidation is often limited to the propositional level, kinds of dualities (e.g., holistic vs. analytic,
in that merely negating a particular proposition (i.e., experiential vs. rational, slow-learning vs. fast-
reversing its truth value) is insufficient to deactivate learning). Third, the APE model remains agnostic
the associative link underlying this proposition. The about whether associative and propositional
process of self-generating propositions can also be processes are supported by distinct mental or neural
uncontrollable, in that such propositions may often be structures. Although the notion of dual-systems may
the “stimulus-driven” result of noticing a systematic provide useful links to basic concepts in
relation between stimuli. neuroscience, we remain skeptical as to whether the
Dual Processes, Dual Systems, or Dual brain can be meaningfully divided into two systems,
Representations? considering that the brain includes multiple
The APE model is a dual-process theory in the specialized regions that mutually interact with each
sense that it distinguishes between two conceptually other to produce a particular behavioral outcome. In
distinct processes on the basis of their operating fact, we doubt whether claiming that the two
principles. Whereas associative processes are defined processes operate in two different systems provides
as the activation of mental associations on the basis any additional prediction over and above the ones
of feature similarity and spatio-temporal contiguity, that are already implied by the dual-process
propositional processes are defined as the validation distinction.
of momentarily activated information on the basis of An important issue in this context concerns the
logical consistency. Deviating from this status of associations and propositions as mental
conceptualization, however, the APE model has entities. Although associative and propositional
sometimes been misinterpreted as a dual-system or processes are conceptually distinct in the sense that
dual-representation theory. Whereas dual-process they are characterized by non-overlapping operating
theories limit their assumptions to the distinction principles, a qualitative distinction between
between two kinds of mental processes (see associations and propositions as conceptually distinct
Gawronski & Creighton, in press), dual-system entities is more difficult to maintain (Gawronski &
theories postulate systematic overlap between Bodenhausen, 2006b). According to the APE model,
multiple distinct dualities (e.g., associative/automatic/ any association turns into a proposition if it is
experiential/holistic/slow-learning vs. propositional/ assigned a truth value. Conversely, any proposition
controlled/rational/analytic/fast-learning) with the depends on activated associations, because we do not
two categories of processes being supported by propose an independent storage of propositions in
different mental or neural structures (e.g., Epstein, memory. Thus, although the APE model
1994; Kahneman, 2003; Lieberman, 2003; Sloman, distinguishes between associative and propositional
1996; Smith & DeCoster, 2000; Strack & Deutsch, processes as two conceptually distinct processes, the
2004). A particular subset of dual-system theories are terms association and proposition should not be
dual-representation theories, which propose the misinterpreted as implying two conceptually distinct
storage of two distinct memory representations of the mental entities.
same object (e.g., Rydell & McConnell, 2006; Another important question in the context of
Wilson, Lindsey, & Schooler, 2000). memory representation concerns the storage of
Although the APE model shares some relational information. For example, the stimulus
assumptions with dual-system and dual- event John is eating the sandwich is inherently
representation theories, it disagrees with both relational, in that it involves a clear structure between
conceptualizations in fundamental ways (Gawronski the actor (John), the described action (eating), and
& Bodenhausen, 2011). First, rejecting the notion of the object of the action (sandwich). Because
dual-representations, the APE model does not assume associative memory has sometimes been described as
a separate storage of associations and propositions in being limited to unstructured, bi-directional links
memory. Instead, all information is assumed to be between concepts that do not include relational
stored in the form of associations, which may or may information (e.g., Lieberman, 2003), proponents of
not pass a propositional assessment of validity. propositional models tend to reject the notion of
Second, the APE model does not assume systematic associative memory on the grounds that memory
overlap between multiple distinct dualities, as representations are inherently relational (e.g.,
proposed by dual-system theories. As we outlined in Mitchell et al., 2009). In response to this criticism, it
detail in this chapter, we do not assume a one-to-one is important to note that structural relations can
mapping between operating principles (associative certainly be accommodated in associative memory
To appear in: J. W. Sherman, B. Gawronski, & Y. Trope (Eds.). Dual-process theories of the social mind. New York: Guilford Press.THE APE MODEL 10
models that are based on distributed networks with pronunciation task. Journal of Experimental
multiple layers of excitatory and inhibitory links Social Psychology, 32, 104-128.
(Smith, 1996). Moreover, as we outlined above, the Blair, I. V., Judd, C. M., & Fallman, J. L. (2004). The
central distinction in the APE model is not between automaticity of race and Afrocentric facial
associative and propositional representations. features in social judgments. Journal of
Instead, we distinguish between associative and Personality and Social Psychology, 87, 763-778.
propositional processes on the basis of their non- Botvinick, M. M., Cohen, J. D., & Carter, C. S.
overlapping operating principles. Of course, any (2004). Conflict monitoring and anterior cingulate
cognitive account has to accommodate the inherently cortex: An update. Trends in Cognitive Sciences,
relational structure of mental representation. 8, 539-546.
However, this issue is fundamentally different from Conner, M., & Armitage, C. J. (2008). Attitudinal
the dual-process distinction in the APE model, which ambivalence. In W. D. Crano & R. Prislin (Eds.),
defines associative and propositional processes in Attitudes and attitude change (pp. 261-286). New
terms of their operating principles (i.e., activation vs. York: Psychology Press.
validation), not in terms of their underlying mental Craik, F. I. M., & Lockhart, R. S. (1972). Levels of
representations. processing: A framework for memory research.
Conclusion Journal of Verbal Learning and Verbal Behavior,
Our main goal in the current chapter was to 11, 671-684.
explain how the distinction between associative and Custers, R., & Aarts, H. (2011). Learning of
propositional processes relates to the distinction predictive relations between events depends on
between automatic and controlled processes that has attention, not on awareness. Consciousness and
guided dual-process theorizing in the last three Cognition, 20, 368-378.
decades. We argued that the associative-propositional De Houwer, J. (2007). A conceptual and theoretical
distinction refers to what a given process is doing analysis of evaluative conditioning. The Spanish
(operating principles), whereas the automatic- Journal of Psychology, 10, 230-241.
controlled distinction refers to empirical claims about De Houwer, J., Thomas, S., & Baeyens, F. (2001)
when that process is operating (operating conditions). Associative learning of likes and dislikes: A
Although the two dichotomies are sometimes review of 25 years of research on human
assumed to overlap, the APE model draws a sharp evaluative conditioning. Psychological Bulletin,
line between operating principles and operating 127, 853-869.
conditions, in that both associative and propositional Deutsch, R., Gawronski, B., & Strack, F. (2006). At
processes have automatic and controlled aspects. the boundaries of automaticity: Negation as
Although the APE model is primarily concerned with reflective operation. Journal of Personality and
the role of associative and propositional processes in Social Psychology, 91, 385-405.
evaluation, its core assumptions are applicable to Devine, P. G. (1989). Stereotypes and prejudice:
both evaluative and non-evaluative information, Their automatic and controlled components.
thereby providing the basis for a general theory of Journal of Personality and Social Psychology, 56,
human thought. 5-18.
References Epstein, S. (1994). Integration of the cognitive and
Baars, B. J. (2002). The conscious access hypothesis: psychodynamic unconscious. American
Origins and recent evidence. Trends in Cognitive Psychologist, 49, 709-724.
Sciences, 6, 47-52. Festinger, L. (1957). A theory of cognitive
Baddeley, A. (2010). Working memory. Current dissonance. Evanston, IL: Row Peterson.
Biology, 20, 136-140 Fiedler, K., & Unkelbach, C. (2011). Evaluative
Balas, R., & Gawronski, B. (2012). On the conditioning depends on higher order encoding
intentional control of conditioned evaluative processes. Cognition and Emotion, 25, 639-656.
responses. Learning and Motivation, 43, 89-98. Field, A. P., & Moore, A. C. (2005). Dissociating the
Bargh, J. A. (1994). The four horsemen of effects of attention and contingency awareness on
automaticity: Awareness, intention, efficiency, evaluative conditioning effects in the visual
and control in social cognition. In R. S. Wyer & paradigm. Cognition and Emotion, 19, 217-243.
T. K. Srull (Eds.), Handbook of social cognition Förderer, S., & Unkelbach, C. (2012). Hating the cute
(pp. 1-40). Hillsdale, NJ: Erlbaum. kitten or loving the aggressive pit-bull: EC
Bargh, J. A., Chaiken, S., Raymond, P., & Hymes, C. effects depend on CS-US relations. Cognition and
(1996). The automatic evaluation effect: Emotion, 26, 534-540.
Unconditional automatic activation with a
To appear in: J. W. Sherman, B. Gawronski, & Y. Trope (Eds.). Dual-process theories of the social mind. New York: Guilford Press.THE APE MODEL 11
Gawronski, B. (in press). Back to the future of Gawronski, B., & Quinn, K. A. (in press). Guilty by
dissonance theory: Cognitive consistency as a mere similarity: Assimilative effects of facial
core motive. Social Cognition. resemblance on automatic evaluation. Journal of
Gawronski, B., Balas, R., & Creighton, L. A. (2012). Experimental Social Psychology.
Associative and propositional processes in Gawronski, B., & Sritharan, R. (2010). Formation,
evaluative conditioning. Manuscript submitted for change, and contextualization of mental
publication. associations: Determinants and principles of
Gawronski, B., & Bodenhausen, G. V. (2006a). variations in implicit measures. In B. Gawronski,
Associative and propositional processes in & B. K. Payne (Eds.), Handbook of implicit social
evaluation: An integrative review of implicit and cognition: Measurement, theory, and applications
explicit attitude change. Psychological Bulletin, (pp. 216-240). New York, NY: Guilford Press.
132, 692-731. Gawronski, B., & Strack, F. (2004). On the
Gawronski, B., & Bodenhausen, G. V. (2006b). propositional nature of cognitive consistency:
Associative and propositional processes in Dissonance changes explicit, but not implicit
evaluation: Conceptual, empirical, and meta- attitudes. Journal of Experimental Social
theoretical issues. Reply to Albarracín, Hart, and Psychology, 40, 535-542.
McCulloch (2006), Kruglanski and Dechesne Gawronski, B., & Walther, E. (2012). What do
(2006), and Petty and Briñol (2006). memory data tell us about the role of contingency
Psychological Bulletin, 132, 745-750. awareness in evaluative conditioning? Journal of
Gawronski, B., & Bodenhausen, G. V. (2007). Experimental Social Psychology, 48, 617-623.
Unraveling the processes underlying evaluation: Gawronski, B., Walther, E., & Blank, H. (2005).
Attitudes from the perspective of the APE Model. Cognitive consistency and the formation of
Social Cognition, 25, 687-717. interpersonal attitudes: Cognitive balance affects
Gawronski, B., & Bodenhausen, G. V. (2009). the encoding of social information. Journal of
Operating principles versus operating conditions Experimental Social Psychology, 41, 618-626.
in the distinction between associative and Greenwald, A. G. (1992). New Look 3: Reclaiming
propositional processes. Behavioral and Brain unconscious cognition. American Psychologist,
Sciences, 32, 207-208. 47, 766-779.
Gawronski, B., & Bodenhausen, G. V. (2011). The Günaydin, G., Zayas, V., Selcuk, E., & Hazan, C.
associative-propositional evaluation model: (2012). I like you but I don’t know why:
Theory, evidence, and open questions. Advances Objective facial resemblance to significant others
in Experimental Social Psychology, 44, 59-127. influences snap judgments. Journal of
Gawronski, B., & Creighton, L. A. (in press). Dual- Experimental Social Psychology, 48, 350–353.
process theories. In D. E. Carlston (Ed.), The Hamilton, D. L., Katz, L. B., & Leirer, V. O. (1980).
Oxford handbook of social cognition. New York: Cognitive representation of personality
Oxford University Press. impression: Organizational processes in first
Gawronski, B., Deutsch, R., Mbirkou, S., Seibt, B., & impression formation. Journal of Personality and
Strack, F. (2008). When “just say no” is not Social Psychology, 39, 1050-1063.
enough: Affirmation versus negation training and Hofmann, W., De Houwer, J., Perugini, M., Baeyens,
the reduction of automatic stereotype activation. F., & Crombez, G. (2010). Evaluative
Journal of Experimental Social Psychology, 44, conditioning in humans: A meta-analysis.
370-377. Psychological Bulletin, 136, 390-421.
Gawronski, B., Hofmann, W., & Wilbur, C. J. Hütter, M., Sweldens, S., Stahl, C., Unkelbach, C., &
(2006). Are “implicit” attitudes unconscious? Klauer, K. C. (2012). Dissociating contingency
Consciousness and Cognition, 15, 485-499. awareness and conditioned attitudes: Evidence of
Gawronski, B., & LeBel, E. P. (2008). Understanding contingency-unaware evaluative conditioning.
patterns of attitude change: When implicit Journal of Experimental Psychology: General,
measures show change, but explicit measures do 141, 539-557.
not. Journal of Experimental Social Psychology, Jacoby, L. L. (1991). A process dissociation
44, 1355-1361. framework: Separating automatic from intentional
Gawronski, B., Peters, K. R., Brochu, P. M., & uses of memory. Journal of Memory and
Strack, F. (2008). Understanding the relations Language, 30, 513–541.
between different forms of racial prejudice: A Jones, E. E., & Gerard, H. B. (1967). Foundations of
cognitive consistency perspective. Personality social psychology. New York: Wiley.
and Social Psychology Bulletin, 34, 648-665.
To appear in: J. W. Sherman, B. Gawronski, & Y. Trope (Eds.). Dual-process theories of the social mind. New York: Guilford Press.You can also read
