TY - JOUR
AU - Reimer, Torsten
AB - Abstract Communication researchers have deplored the absence of theory-driven criteria of argument quality in persuasion. Probabilistic Persuasion Theory (PPT) aims to offer theoretical criteria which define the quality of arguments, a priori. Past experimental research tested the validity and distinctiveness of cues as argument quality criteria. Drawing on semantic network theory, the present research introduces two new criteria of argument quality—attribute degree centrality and attribute tie strength. Analyses of semantic network data collected on over-the-ear headphones attributes from the environment and cognitive representations are reported. The analysis demonstrates how degree centrality and tie strength can be measured and how semantic networks can be used to distinguish strong and weak arguments. Introduction A key virtue of the human mind is the ability to make adaptive decisions under conditions of time and information scarcity (Hoffrage & Reimer, 2004; Reimer & Rieskamp, 2007). One important way an organism can adapt to its environment is by forming inferences that go beyond the information that is given in a certain situation (Smith & Kosslyn, 2007). For illustration, suppose Sally needs to choose between two physicians at different primary care offices. She calls both and is told a nurse would respond by the end of the day. Incidentally, one nurse calls back and the other does not. Without further investigation, Sally goes with the physician whose nurse called her back. While the nurse calling back does not provide direct evidence about the physician per se, this piece of information may enable Sally to infer relevant information about the physician and the primary care office. For example, Sally may infer that the office that called back is probably more reliable and cares more about their patients than the office that did not call back. These inferences are adaptive to the extent that they reflect actual relationships among the involved attributes in the real world. Moreover, consider a situation where one wants to persuade Sally to go with one office over the other. Would embedding the cue “call’s back on the same day” in a message permit Sally to infer other positive attributes regarding the physician and hence be persuaded to choose the primary care office advocated by this message? The present research seeks an answer to this question by analyzing the underlying semantic networks that describe the actual and perceived connections among relevant attributes. Based on the assumption that knowledge about semantic networks can be utilized for the generation of arguments and persuasive messages, the introduced approach offers theory-based criteria of argument quality and outlines novel ideas about persuasion effects that can be tested in future research. In search of argument quality criteria Aristotelian rhetorical theory differentiates three means of persuasion corresponding to the existence of a speaker, audience, and argument. Persuasion may occur through a speaker’s character (ethos), an audience’s emotional state (pathos), and the argument itself (logos). A persistent pursuit in the contemporary study of argumentation and persuasion is to specify message properties which theoretically define the quality of arguments. Communication researchers have lamented the lack of argumentation theory in persuasion (O’Keefe, 2003; Reimer, 1997, 2003; Reimer et al., 2012; Stiff, 1986). Without theoretical grounds for defining the quality of arguments, no coherent explanation for the persuasive effects of arguments is possible (O’Keefe, 2003; Reimer et al., 2012). Effects-based approaches to argument quality have provided several indices of perceived argument strength (e.g., perceived believability, perceived convincingness; see Zhao, Strasser, Cappella, Lerman, & Fishbein, 2011, for a discussion). These approaches, however, have avoided identifying theory-based dimensions of argument quality. In this tradition, argument quality is purely operationalized through receiver evaluations—an argument is assumed to be as strong as it is perceived without providing any theoretical criteria (O’Keefe, 2003; Reimer et al., 2012). As a consequence, these approaches do not provide any guidance for the development of strong or weak arguments. Increased comprehension and explanation of persuasion processes requires developing and testing a priori dimensions of the quality of arguments. More recently, several research groups have started to take critical steps, advancing theory-based approaches to the study of argument quality (e.g., Hahn & Oaksford, 2007; Hoeken, Timmers, & Schellens, 2012; Hoeken, Šorm, & Schellens, 2014; Reimer, Raab, & Russell, 2017; Reimer et al., 2012; also see Mercier & Sperber, 2011). For example, prior experimental research indicates that enthymematic validity may function as a criterion of argument quality (Reimer, 2003; Reimer et al., 2017; also see Trouche, Johansson, Hall, & Mercier, 2015). Arguments, in this research, were operationalized as high in quality if completed by a valid presupposition. Valid enthymemes, defined within Aristotelian logic, were found to be higher in perceived argument quality than arguments based on enthymemes lacking validity (Reimer et al., 2017). Research capitalizing on rational decision theory suggests the desirability and likelihood of a consequence as criteria of the quality of arguments (Hoeken et al., 2012, 2014; Hornikx & Hoeken, 2007). Building on argumentation schemes, argument quality has also been defined in terms of resistance to critical questioning and scrutiny (Hoeken et al., 2012, 2014; Walton, 2005). Using Bayesian probability theory, prior research has developed and tested subjective probabilities related to persuasive message content as theory-driven criteria of the goodness of arguments (Hahn & Oaksford, 2006; 2007; Harris, Hsu, & Madsen, 2012). In line with this body of research, a novel source for the derivation of argument quality criteria has recently been proposed, namely the structure of the information environment (Reimer et al., 2012). A cue’s ecological validity and distinctiveness determined by a reference class have been developed and tested as theory-based indicators of the quality of arguments (Reimer et al., 2012; Russell, Reimer, & Hertwig, 2014). The present study aims to contribute to, and extend, this research agenda by offering two independent criteria drawn from the semantic networks literature, attribute degree centrality (i.e., an attribute’s number of direct connections to other attributes in a semantic network) and attribute tie strength (i.e., an attribute’s strength of direct connection to other attributes in a semantic network). In order to introduce attribute degree centrality and attribute tie strength as criteria of argument quality, we first place the present research in the larger conceptual framework provided by Probabilistic Persuasion Theory (PPT). In a second step, we provide a brief example of how attributes varying along both dimensions can be embedded in pragmatic arguments (see O’Keefe, 2013, for an overview of pragmatic argumentation research). Finally, in a third step, we make predictions concerning the message effects of attribute degree centrality and tie strength on a number of persuasion and decision-making outcomes to facilitate future research. Probabilistic Persuasion Theory and semantic networks Different from theories of argument quality derived from Bayesian (e.g., Hahn & Oaksford, 2006, 2007) or expected utility (e.g., Hoeken et al., 2012) models of judgment and decision-making, PPT capitalizes on research suggesting that deciders often use noncompensatory, frugal heuristics such as Take-The-Best when forming decisions (see Gigerenzer, Czerlinski, & Martignon, 1999; Gigerenzer, Todd, & the ABC Research Group, 1999; Hoffrage & Reimer, 2004). PPT rests on Egon Brunswik’s probabilistic functionalism and the Brunswikian lens model (Reimer et al., 2012). Probabilistic functionalism envisages organisms as using cues to predict outcomes in an intrinsically uncertain and probabilistic social environment (Brunswik, 1935, 1955; Reimer et al., 2012). Brunswik conceptualized judgment processes as a “coming-to-terms” with environment structure (Brunswik, 1944; Tolman & Brunswik, 1935). Organisms make judgments based on proximal cues which are probabilistically related to a distal criterion (Tolman & Brunswik, 1935). The extent to which a cue predicts a distal criterion is an index of its ecological validity (Reimer et al., 2012). PPT builds on the lens model and the concept of ecological validity by suggesting cue dimensions can provide theoretical criteria of argument quality. PPT assumes the functional equivalence of cues and arguments (Kruglanski & Thompson, 1999; Reimer et al., 2012). With respect to both function and mode of processing, cues and arguments are indistinguishable within the persuasion context (Kruglanski & Thompson, 1999; Reimer et al., 2012). Cues may pertain to external object properties, external or perceived attributes of a speaker or context, or a listener’s knowledge state (Reimer et al., 2012). A cue’s correlation with environmental criteria (i.e., cue validity), or number of other objects for which that cue is absent in a reference class (i.e., cue distinctiveness), are cue characteristics that can be used to define the quality of arguments. Though the actual quality of arguments in this approach can be determined independent of the perception of message receivers, perceptions of argument quality appear to be remarkably accurate. Research has shown that, typically, receivers’ perceptions of cues validities are systematically related to their actual validities (Katsikopoulos, Schooler, & Hertwig, 2010). In general, as people come to terms with the environment by gaining experience with relevant reference classes, their perception of cue validities will more closely correspond to ecological validities embedded in the information environment (Gigerenzer, Hoffrage, & Kleinbölting, 1991; Katsikopoulos et al., 2010). PPT assumes that deciders capitalize upon these good intuitions about the relationship between cues and criterions when constructing mental models of a decision task. Research in decision-making has claimed that when people face a choice task under uncertainty (for example, the task to choose between two primary care offices) they construct a probabilistic mental model (PMM) (Gigerenzer et al., 1991). PMMs are devices permitting induction based on co-occurrence frequencies stored in memory. For instance, in the primary care office example, a decider might think that reliability is a very relevant and valid cue that is systematically related to a variety of key performance measures of primary care offices. In this case, one would expect that the reliability cue is part of, and will be considered within, a decider’s constructed mental model. PMMs are thought to exploit cue-criterion structures based on relationships among cues and criteria in the information environment (Brunswik, 1955; Gigerenzer et al., 1991). PPT outlines basic processes that can be used to influence a decider based on characteristics of the decider’s mental model. For example, if a persuader knows how deciders typically rank order the importance of attributes or cues in a decision-making situation, the persuader is able to tailor messages to a decider’s mental model by appealing to the attribute likely to be perceived as highest in importance (Reimer et al., 2012). This may provide key editorial criteria for argument choice in a persuasion context (Hample, 1990). PPT describes central features of the mental models that are used by deciders and offers strategies that can be used by a persuader who seeks to influence by capitalizing on, or modifying, the mental model of a decider. Importantly, a decider’s mental model does not affect actual argument quality (e.g., cue-criterion correlations in the environment) determined by the reference class but whether an argument is likely to be persuasive. The current approach builds upon the basic assumption of the PMM and PPT approaches that people construct mental models when they form a decision (Gigerenzer & Goldstein, 1996; Gigerenzer et al., 1991). Research on mental models in decision-making has systematically described major relationships among cues in the environment as well as their perception (such as the correlation among cues). Research on semantic networks promises to provide a fruitful approach to extend this approach by describing network-relationships amongst cues, offering a rich tool to explore the role of central characteristics of mental models constructed by deciders. Network models suggest semantic memory can be represented as a set of units (or nodes) connected to each other in a network (Chang, 1986; Collins & Loftus, 1975; Collins & Quillian, 1969). This network structure of semantic memory may impact the accessibility and utilization of specific units of information in a decision-making context. In this way, semantic network models may assist with understanding how relevant cues are connected to each other and how these connections are employed in a decision-making and persuasion situation. Semantic network theory aids with grasping how degree centrality and tie strength can be used to conceptualize the structure of environmentally-based cue information. This information is stored in long-term memory and can be retrieved in situations in which a decider constructs a PMM. On the most basic level, semantic networks can be defined as a set of nodes with connections amongst these nodes. Each node is connected to a specific number of other nodes with each connection possessing a specific strength. Though semantic network nodes can refer to all four categories of cues (i.e., object, speaker, listener, and context) proposed by PPT, the present research focuses on cues reflecting external attributes of objects. Object attributes can be positive, neutral, or negative. For the sake of simplicity, semantic network nodes in the current project refer to positive attributes. Not capitalizing on people’s perception of attribute valence, we conceptualize positive attributes as tangible features of an object (e.g., a primary care office with high performance ratings). The semantic network describes the connections among those positive attributes (e.g., positive attributes of a primary care office) given a specified reference class of objects (e.g., all offices in a town). The current approach focuses on two key characteristics of attributes: the number of other positive attributes in the network an attribute is connected with (i.e., degree centrality) and the strength of those connections (i.e., tie strength). As in the case of validities, an external reference class can determine degree centrality and tie strength scores of all attributes. As a consequence, degree centrality and tie strength can be determined independent of the perception of message receivers. Given a specific task structure, the higher the positive attribute’s centrality, the higher the actual quality of the argument utilizing this positive attribute. The extent to which a decider is persuaded by a highly central attribute (or perceives it as high in quality) should depend on whether a decider’s cognitive representation sufficiently matches the environmentally-based semantic network structure. If semantic network structure is cognitively represented and stored in long-term memory, we submit the number and strength dimensions of attributes may facilitate inferential processes of the organism, consequently impacting persuasion and decision-making. Specifically, the quantity and speed of information retrieval made possible by high in degree centrality and tie strength attributes may affect the persuasiveness of communicators by activating in the mind of decision-makers more decision-relevant attributes (and activate them more quickly) than low in degree centrality and tie strength attributes. Following the tradition in social network analyses (Fink, High, & Smith, 2015; Lee & Monge, 2011; Monge & Contractor, 2003; Wasserman & Faust, 1994), the current project aims to utilize the specific measures and techniques developed to study the number and strength of connections amongst social units in a persuasion context. Semantic network theory and persuasion Communication researchers have a long-standing interest in understanding how cognitive structure may influence persuasion processes and outcomes (See Fink, Monahan, & Kaplowitz, 1989; Woelfel & Fink, 1980; Woelfel, Gillham, Cody, & Holmes, 1980). Consistent with this tradition, the present research suggests semantic network theory may assist with providing theory-based criteria of argument quality. Semantic network theories have typically conceptualized nodes as representations of specifiable units of category information (Tulving, 1972). Originally discussed within categorization research, Tulving (1972) distinguished semantic and episodic memory. Episodic memory refers to storage of personal memories with specifiable dates and times. Conversely, semantic memory refers to storage of organized networks of factual knowledge, concepts, and concept properties. Within the context of semantic network research, relationships have traditionally referred to either logical or associative relationships between object attributes, features, or properties. Collins and Loftus (1975) developed the Spreading Activation Model (SAM) to explain and integrate the findings of semantic memory research. SAM would become one of the most popular and often cited network models of semantic memory. Nodes in the SAM network are concepts with links indicating logical relationships, typicality gradients, and degrees of association. The model assumes that when a concept node is being activated, activation spreads to concepts that are connected to it (Collins & Loftus, 1975). The more connections two concepts have in common and the stronger these connections, the more related they are in a semantic memory network (Collins & Loftus, 1975). The more related one concept is to others in semantic memory, the higher the likelihood that the processing of this concept will give rise to the processing of others (Collins & Loftus, 1975). The underlying premise of the classic spreading activation perspective is that retrieval of semantic information is ultimately a function of the number and strength of connections in a semantic memory network. SAM has a means of explaining most of the major semantic memory effects including but not limited to the category size effect (verification is faster for small category pairs than large category pairs), the familiarity effect (category familiarity increases category size effect), typicality (verification is faster for typical pairs than atypical pairs), and the false production frequency effect (frequently associated pairs are verified faster than unfrequently associated pairs; see Chang, 1986; Collins & Quillian, 1969; Loftus, 1973). How exactly can the findings and explanatory mechanisms related to the retrieval of semantic information be applied to a persuasion context? A simple example may help with explaining how this is possible. Semantic networks in persuasion: an example of pragmatic argumentation Semantic networks offer fertile ground for the description and examination of the connections between cues in the environment and how this structure provides criteria of argument quality. In addition to offering argument quality criteria, we also suggest the perceived number (i.e., degree centrality) and strength (i.e., tie strength) of connections to other cues can affect how persuasive the respective cues are when used as arguments in a persuasion context. The present research aims to demonstrate how semantic networks can be utilized in the context of pragmatic argumentation (see O’Keefe, 2013) by conceptualizing semantic network nodes as external, positive attributes of an object. Pragmatic arguments (or arguments from consequence) are used in many persuasion studies and refer to arguments which support the choice of a specific alternative by citing the desirable or undesirable consequences of the advocated action (O’Keefe, 2013; Walton, 1996; Perelman, 1959). Not all arguments are pragmatic arguments—pragmatic arguments are arguments that refer to the consequences of following an argumentation. While argumentation researchers have examined the relative persuasiveness of different argument types including, but not limited to, arguments from analogy, arguments from authority, and arguments from example (Hoeken et al., 2014), the consequence-based argument has received the most thorough experimental consideration in persuasion research (see O’Keefe, 2013, for an overview). In the following example, we illustrate how attribute degree centrality and tie strength can be applied to the context of pragmatic argumentation. Consider the following scenario: Steve works at a retail store that sells over-the-ear headphones and is attempting to sell a specific headphone to a customer. The customer does not have much time and is looking to buy the headphone with the most positive attributes. Steve can only choose one headphone attribute to embed in a persuasive message. Under the assumption that the customer is looking for the headphone with the most positive attributes, which argument, A or B, is better? Which argument will be more persuasive? Argument A: Purchase the Loneg JBT-1380 because it has Sound Isolating Technology. Argument B: Purchase the Loneg JBT-1380 because it has a Detachable Cable. Independent of the customer’s perception, given the task to pick the headphone with the most positive attributes, the objectively better argument is the one that uses an attribute connected to more positive headphone attributes in the external world. Without further cue knowledge, the best argument for choosing the Loneg JBT-1380 is the argument whose attribute is actually connected to more positive attributes. On the other hand, we predict—ceteris paribus—that the more persuasive argument will depend on the customer’s cognitive representation of the semantic network structure. If Steve knows how people typically rank order positive headphone attributes in terms of number and strength of connections to other positive attributes, he can tailor his message to the mental model of the customer by appealing to the positive attribute likely to be perceived as highest in degree centrality and tie strength. Figure 1 provides an example of how that structure may appear. If these network configurations exist in the external environment, Sound Isolating Technology is a higher-quality argument in favor of choosing the JBT-1380. It is strongly connected to three other positive attributes, whereas Detachable Cable is only weakly associated with one other positive attribute. If people’s cognitive representations are adequately isomorphic with this environmental structure, then Steve should choose Sound Isolating Technology to persuade his customer because it is also likely that this attribute will be perceived as high (relative to Detachable Cable) in degree centrality and tie strength. Degree centrality and tie strength are both dimensions of positive attributes which may define the actual quality of arguments and, with sufficient cognitive adaptation to environmental structure, may influence the persuasiveness of arguments through spreading activation processes. Figure 1 View largeDownload slide Example of how semantic network structure can be applied to the persuasion context. Thick lines refer to high tie strength attribute relations, thin lines refer to low tie strength attribute relations, and dashed lines indicate a lack of relationship amongst attributes. Arrows indicate spreading activation processes. Sound Isolating Technology is a high degree centrality/high tie strength attribute. Detachable Cable is a low degree centrality/low tie strength attribute. Figure 1 View largeDownload slide Example of how semantic network structure can be applied to the persuasion context. Thick lines refer to high tie strength attribute relations, thin lines refer to low tie strength attribute relations, and dashed lines indicate a lack of relationship amongst attributes. Arrows indicate spreading activation processes. Sound Isolating Technology is a high degree centrality/high tie strength attribute. Detachable Cable is a low degree centrality/low tie strength attribute. Attribute degree centrality and tie strength Attribute degree centrality indicates the extent of an attribute’s direct connections to other attributes in a semantic network. Often in the process of decision-making, one must retrieve semantic information from memory (Peter, Olson, & Grunert, 1999). This information has a specific organization and structure that is usefully represented as a set of object attributes connected to a specific number of other attributes (Cowley & Mitchell, 2003; Grebitus & Bruhn, 2006; Jonassen, Beissner, & Yacci, 1993). The information that can be retrieved depends on the organization of these attributes, one dimension of which is the number of connections that an attribute has with others. When an attribute is high as opposed to low in attribute degree centrality (i.e., has many connections) it should theoretically increase the amount of accessible information (Chang, 1986; Collins & Loftus, 1975). When embedded in arguments, we predict an argument based on attributes high in degree centrality will be a stronger than one based on attributes low in degree centrality. Attribute tie strength, conversely, indicates an attribute’s average connection strength to other attributes in a semantic network to which it is directly connected. Empirical evidence supports the view that the speed of information retrieval impacts decision-making under uncertainty (Hertwig, Herzog, Schooler, & Reimer, 2008). This finding corroborates classic semantic network research that showed that the more strongly related two concepts are in memory, the faster one concept can be retrieved from the presence of the other (Chang, 1986; Collins & Loftus, 1975; Collins & Quillian, 1969). We predict that when embedded in arguments, attributes that are high in tie strength will be stronger arguments and yield shorter choice latencies than arguments with attributes that are low in tie strength. Along with impacting choice, we also predict that attribute degree centrality and tie strength will increase participant choice confidence to the extent that connected attributes support a decision. Research suggests two primary determinants of subjective confidence in judgment relevant to degree centrality and tie strength: the amount of evidence that comes to mind and the speed by which it comes to mind. A body of research on judgment confidence suggests that, in general, increasing the amount of supporting evidence retrieved for an answer to a judgment item will increase confidence in an answer’s accuracy (Glucksberg & McCloskey, 1981; Graesser & Hemphill, 1991; Koriat, Lichtenstein, & Fischhoff, 1980). One useful way to conceptualize the impact of an attribute’s degree centrality on the mind is to think of it as a stimulus providing access to a larger quantity of evidence related to an alternative. To the extent that degree centrality increases the amount of other positive attributes a specific attribute activates in a semantic network we predict it will increase the confidence participants have in their choices. If an attribute triggers other attributes, higher as opposed to lower tie strength should also increase participant choice confidence. Research suggests that confidence in judgment accuracy is partly determined by how quickly answers come to mind, irrespective of the actual accuracy of the answers (Nelson & Narens, 1990). Other evidence has indicated that the more quickly participants retrieve information about a text the higher the confidence in their ability to answer items about that text (Begg, Duft, Lalonde, Melnick, & Sanvito, 1989; Morris, 1990). Consistent with this body of research, Kelley and Lindsay (1993) directly compared the speed by which answers to general knowledge items come to mind for 17 participants and found that it increased confidence in given answers. As tie strength should theoretically increase the speed of attribute activation, we predict it will augment participant choice confidence. We also predict that attribute degree centrality and tie strength will increase positive evaluations of message. Arguments which activate more (as opposed to less) supporting evidence for an alternative should influence argument quality ratings. Recent research suggests that the level of ease or difficulty in retrieving information is related to stimulus liking (Whittlesea, 1993; Winkielman & Cacioppo, 2001). Stimulus characteristics such as repeated stimulus exposure, pattern goodness, averageness, and prototypicality may facilitate a receiver’s information processing, yielding both efficient processing and increased stimulus preference ratings (Henley, Horsfall, & De Soto, 1969; Winkielman & Cacioppo, 2001). Winkielman and Cacioppo (2001) asked 16 participants to watch a series of pictures of everyday objects that were affectively neutral. The researcher’s manipulated picture processing ease through subliminal presentation of contour primes which either matched or mismatched the target. To examine the effect of processing ease on stimulus liking, participant facial electromyograms (EMG) and self-reports of stimulus liking were gathered and examined. As expected, targets that were easier to process tended to occasion physiological responses and self-reports that indicated positive affect. Consistent with this research, we predict that in our context, in which participants face the task to select an object that has as many attributes as possible and in which all attributes are positive, attribute degree centrality and tie strength will affect the speed by which attributes not present to the decision-maker are activated, increasing positive ratings of the corresponding message. Method As a first step in examining whether degree centrality and tie strength offer theory-driven criteria of argument quality, the present research explores whether and to what extent semantic network structure exists in the environment and the human mind. To this end, we generated an environmentally-based semantic network that describes attribute interconnection in a real-world example—over-the-ear headphones. Over-the-ear headphones in this research refer to headphones that sit either on, over, or around the ear in contrast to “earbuds” which are placed inside the ear. We chose this example since it represents an area in which most people have some experience and knowledge about object features and attributes without feeling very strongly attached to specific attributes. To study how domain-specific attributes are connected in people’s minds and the extent to which cognitive representations match information structure in the external environment, we used concept-mapping methods. Studying the environment enables identification of effects-independent criteria of argument quality. Analyses of cognitive representations of semantic network structure are based on specification of a theory-based mediating mechanism (in the sense of O’Keefe, 2003) proposed to influence persuasion outcomes. Four fundamental relations can exist between a specific attribute co-occurrence configuration in the environment (e.g., Sound Isolating Technology is strongly connected to many other attributes) and a specific cognitive representation. First, the configuration (e.g., a strong correlation) exists in the environment and the cognitive representation matches this structure. Second, the configuration exists in the environment but the cognitive representation does not match this structure. Third, the configuration does not exist in the environment but it is cognitively represented. Fourth and finally, the configuration does not exist in the environment nor is it cognitively represented. With the objective of examining which of these four relationships describe the semantic networks of over-the-ear headphones, we measured headphone attributes derived from the environment and the cognitive representations of participants. Three novel questions guided the analysis of the data. First, with respect to the environment, do headphone attributes differ along the two dimensions of degree centrality and tie strength and are these two dimensions correlated? Second, with regard to participants’ cognitive representations, are participants sensitive to variation in the degree centrality and tie strength of attributes derived from the environment? Third, and finally, what are the similarities and differences across the cognitive representations of participants with respect to how accurately they perceived network structure? Participants and design Environment Data on over-the-ear headphones were collected by recording spec data from a commercial website of a headphone retailer, Best Buy. This store was the most successful retail seller of over-the-ear headphones in 2013. Forty headphone attributes were derived from 50 headphone spec lists gathered from Best Buy’s website. Fifty 40 × 40, co-occurrence matrices were produced from attribute data and one aggregate matrix was produced from the 50 matrices. Cognitive representations Collecting cognitive representation data involved 17 participants from a university in the Midwestern United States (eight males, nine females; mean age, M = 18.9; SD = 1.10). Seven hundred eighty items, administered in a random order, yielded 17 40 × 40 co-occurrence matrices. Procedure Environment The first objective involved generating a semantic network from product attribute data sampled from the environment. To this end, over-the-ear headphone attribute data was gathered from the 50 most popular over-the-ear headphones listed on Best Buy’s retail website. Retail websites typically list product attributes on spec pages along with attribute subsets chosen for product advertising. We systematically recorded over-the-ear headphone attributes describing the 50 most popular over-the-ear headphones listed on Best Buy’s retail website. Data collection yielded 40 attributes. Next we recorded, for all attributes, whether or not any particular over-the-ear headphone (out of the 50) possessed a specific attribute. We then transformed all attributes into a binary format by means of a median split procedure. After the attribute data were recorded and all attributes were transformed into binary format, we produced 50 40 × 40 co-occurrence matrices representing each of the over-the-ear headphones. The term co-occurrence refers to whether, on any particular over-the-ear headphone H, two attributes i and j occur together. For any cell in a specific co-occurrence matrix, if i and j occurred together on H, the cell received a value of 1. If i and j did not occur together, the cell received a value of 0. Though all attribute information was recorded for every headphone across the 50, not every attribute spec list for every headphone contained information on every attribute. If an attribute was not on the spec list, this attribute’s connections to all other attributes received a zero. After producing all 50 matrices, we summed across all 50 to produce one matrix representing the total co-occurrence data on all 40 attributes. A critical motive for producing the aggregate matrix was investigation of whether semantic network structure was inherent in this data and whether attributes varied in terms of their degree centrality and tie strength. A second aim for producing the matrix was examination of whether people perceive this semantic network structure in the environment and are sensitive to variation in degree centrality and tie strength. The second objective required data collected on how individuals perceive the relationships amongst the 40 attributes describing the 50 over-the-ear headphones in the reference class. Cognitive representations We used concept-mapping methodology to explore how people perceive semantic network structure derived from the environment. An established methodological paradigm, concept mapping permits production and analysis of cognitive representations people possess concerning relationships amongst over-the-ear headphone attributes (see Grebitus and Bruhn, 2006, for an example of concept mapping applied to perceived pork quality). Using the same 40 attributes gathered from the environment, we asked 17 participants to report to what extent (i.e., if and how often) any two attributes, i and j, appeared across 50 over-the-ear headphones they may encounter when shopping at a retail store. Example: How many (out of 50) different pairs of over-the-ear headphones have Noise Cancelling Technology and Good Magnet Type (Neodymium)? Participants responded to items of this type on a continuous scale ranging between 0 and 50 by sliding dials specifying on how many over-the-ear headphones they believed particular pairs of attributes occurred together. Seven hundred eighty randomly ordered items yielded a 40 × 40 matrix from participant responses. Across all items, we counterbalanced attribute sequence within items such that half of the attributes appear first and second in sequence as often as the other half of attributes. Overall, this procedure enabled the generation of 17 unique cognitive representations to be generated. Measures Attribute degree centrality Degree centrality indicates the number of direct connections attributes have with other attributes in the semantic network. An attribute is high in degree centrality when, compared with other attributes, it is connected with a larger number of attributes. We defined degree centrality of an attribute Pk as the number of adjacent attributes directly connected to Pk (see Freeman, 1979). Across the 50 over-the-ear headphones, an attribute’s degree centrality was determined by counting with how many other attributes an attribute co-occurred at least once. Standardization of attribute degree centrality scores involved dividing each score by the total number of possible connections they could have with other attributes in the network. For example, if attribute Pk was connected to two other attributes out of a total of three possible attributes the attribute degree centrality of Pk would be .66. Attribute degree centrality does not take into account how often—that is, for how many headphones—two attributes co-occur. The frequency of co-occurrence determines attribute tie strength. Attribute tie strength Attribute tie strength refers to the frequency by which a pair-wise connection between attributes appears across a set of objects in a reference class. Across the 50 headphones, attribute tie strength was determined by the average number of headphones for which a relation exists. By dividing this value by the total number of headphones in the reference class, attribute tie strength values were standardized such that they varied between 0 and 1. For example, if an attribute Pk was connected to three other attributes and these connections occurred, in total, 15 times across a reference class of six headphones, the attribute tie strength of Pk would be .83. In this instance, the tie strength score (.83) is computed by dividing the total number of times the focal attribute’s direct connections occurred, 15, by its number of direct connections, 3; and then dividing that quotient, 5, by the total number of headphones, 6. Weighted degree centrality Weighted degree centrality of an attribute refers to the overall number of connections that this attribute has with other attributes across a reference class. By dividing this value by the total number of headphones in the reference class, multiplied by the total number of possible connections, we standardized weighted degree centrality such that it varied between 0 and 1. For example, if an attribute Pk was connected to two other attributes (out of four possible attributes) and these connections occurred six times, in total, across a reference class of five different headphones, the weighted degree centrality of Pk would be .30. In this case, we computed the weighted degree centrality score (.30) by dividing the total number of times the focal attribute’s direct connections occurred, 6, by the product of the total number of headphones, 5, and the total number of attributes, 4. Results Attribute co-occurrence in the environment The degree centrality scores ranged from .08 to .85 suggesting that attributes do indeed differ in terms of degree centrality. The mean of degree centrality scores across all attributes was .44 (SD = .24). The attribute with the highest degree centrality of .85, Connector 3.5 mm (1/8″) Mini Phone Jack, co-occurred with 33 other attributes at least once. Conversely, the attribute with the lowest degree centrality of .08, Automatic Shutoff Technology, co-occurred with only one other attribute. Attributes also differed with respect to tie strength, though tie strength variation was much less substantive than degree centrality variation. The highest tie strength score for an attribute was .10 and the lowest tie strength score was .02. The mean tie strength score across all attributes was .04 (SD = .02). This suggests that some attributes have stronger connections than others to other attributes across the reference class of 50 headphones. The attribute with the highest tie strength of .10 was the attribute that also had the highest degree centrality: The attribute Connector: 3.5 mm (1/8″) Mini Phone Jack had connections which occurred on an average of about five different headphones. Conversely, the attribute with the lowest tie strength of .02, Wireless Transmitter, had connections which occurred on an average of only one headphone. Less variation in tie strength (when compared with variation in degree centrality) may be explained by the large number of headphones in the reference class (i.e., 50 different headphones) for which connections did not occur. Although some attributes were connected to many other attributes and had a high degree centrality, most of the time, the connections occurred only within a few of the 50 headphones yielding low overall tie strength. In the environment, attributes not only varied in terms of number and strength of connections to other attributes, but the degree centrality and tie strength scores of attributes in the environment were highly correlated with each other, r(40) = .86, p < .001. This substantial correlation suggests that attributes that were high in degree centrality tended to also be high in tie strength and vice versa. After establishing that semantic network structure existed in the environment, the further question concerned whether and to what extent this structure, inherent in the environment, is perceived by participants. Cognitive sensitivity to attribute co-occurrence A primary goal of the present research was to examine whether semantic network structure exists in the environment such that this structure may provide argument quality criteria, a priori. A second related objective was to facilitate research by enabling choice of attributes high and low in centrality across both the environment and cognitive representations. Taking into account the strong, positive correlation between degree centrality and tie strength scores in the environment, we focused our analysis on comparing the weighted degree centrality scores in the environment with the weighted degree centrality scores for cognitive representations as weighted degree centrality integrates the number and strength of connections of an attribute. The analysis had three major components. In a first step, we examined how similar people’s cognitive representations were in terms of their weighted degree centrality estimates for all 40 attributes. In a second step, we examined the similarity between the weighted degree centrality scores in the environment and the average weighted degree centrality scores in the mind. In a third and final step, we compared each individual person’s weighted degree centrality scores for the 40 attributes with the weighted degree centrality scores in the environment. Participant’s cognitive representations were relatively homogenous in terms of how they scored the weighted degree centrality of attributes. An intraclass correlation coefficient of .69 indicated a moderate agreement across participants with respect to perception of an attribute’s weighted degree centrality. This homogeneity across representations was particularly salient on the extremes of the distribution—participants greatly agreed in their perception of the most and least central attributes. We ranked, for each participant, all 40 attributes from largest to smallest with regard to weighted degree centrality. Next, we considered the similarity across cognitive representations with respect to the top 10 and bottom 10 attributes. The same nine attributes appeared among the top 10 in terms of weighted degree centrality across a substantial proportion of the 17 cognitive representations. The attribute (out of the nine) that appeared the least number of times was found in the top 10 across 41% of cognitive representations and the attribute that appeared the most number of times was found in the top 10 across 52% of cognitive representations. Similarly, seven attributes were among the bottom 10 in terms of weighted degree centrality across a considerable proportion of cognitive representations. The attribute (out of the seven) with the least agreement was found in the bottom 10 across 41% of cognitive representations. The attribute with the greatest agreement was found in the bottom 10 across 76% of cognitive representations. This finding suggests that there was moderate similarity across cognitive representations with respect to which attributes were in the top and bottom 10 in terms of weighted degree centrality. On other hand, when comparing the attributes in the bottom 10 alone and top 10 alone there was a greater difference between the lowest and highest values in the lowest 10 than in the highest. In addition to similarity across representations we also examined the extent to which, taken as an average, people’s representations matched the external environment. Overall, people’s cognitive representations match the environment well by showing a substantial positive correlation in weighted degree centrality. This result was obtained by computing a standardized, weighted degree centrality score for each attribute across all 17 cognitive representations. These scores were summed and then divided by 17 to produce a standardized, average weighted degree centrality score for each of the 40 attributes. Once the average weighted degree centrality scores were generated, we examined whether people perceived the same attributes to be high or low in weighted degree centrality that were actually high or low in the environment. The weighted degree centrality scores in the environment and the average weighted degree centrality scores across cognitive representations were moderately correlated r(40) = .44, p < .01. This finding indicates that participants were, on average, sensitive to the weighted degree centrality structure in the environment. Considerable overlap was also revealed when we examined the extremes of the weighted degree centrality scores across both the mind and the environment. We first divided the two sets of rankings into a top 10 and a bottom 10 in weighted degree centrality. Out of the top 10 attributes that were highest in weighted degree centrality in the environment, seven of these attributes were also perceived to be in the top 10 in weighted degree centrality by participants. Out of the bottom 10 attributes that were lowest in weighted degree centrality in the environment, four were also perceived to be in the bottom 10 in weighted degree centrality by participants. In addition to comparing average weighted degree centrality scores across cognitive representations with weighted degree centrality scores in the environment, we also considered each cognitive representation individually in terms of accuracy of semantic network structure perception. Some people were more accurate in how they perceived the weighted degree centrality of attributes in the environment. We first computed Pearson correlations between each person’s weighted degree centrality scores for all 40 attributes and the weighted degree centrality scores for all 40 attributes in the environment. Following Corey, Dunlap, and Burke (1998) we then transformed the individual r values using the Fisher’s z transformation, averaged the z-values, and then performed a z to r value conversion. The average correlation was .17 (SD = .33). Second, we rank ordered participants in terms of their correlation with the environment. This enabled the segmentation of the 17 cognitive representations into a top five, middle seven, and bottom five in terms of structure perception accuracy. For the top five most accurate participants in terms of correlation with the environment, the average correlation was .47 (SD = .11). In contrast, the middle seven and lowest five had average correlations with the environment of .22 (SD = .11) and −.23 (SD = .26), respectively. What these results indicate is that there are systematic differences in how accurate people perceive the weighted degree centrality of attributes in the environment. Discussion Persuasive messages may include arguments varying along a number of dimensions including but not limited to length, order, number, and content. A singularly important way in which an argument may vary is with respect to its quality or goodness (O’Keefe & Jackson, 1995). An argument may be good or poor. Though researchers have operationalized argument quality in several ways (O’Keefe & Jackson, 1995; Zhao et al., 2011), a piercing criticism has concerned the lack of theory-driven criteria for what constitutes a good argument (Boller, Sway, & Munch, 1990; O’Keefe, 2003; Reimer et al., 2012). Instead of specifying actual message features which theoretically define the quality of arguments, researchers have often adopted an empirically driven way of operationalizing argument quality which neglects exploration of intrinsic message properties responsible for specific effects (O’Keefe, 2003; Reimer et al., 2012). In order to address this lack, PPT aims to provide theory-driven dimensions of argument quality knowable a priori. Different from research on persuasion effects that operationalize argument quality merely empirically without providing theory-based criteria (see Reimer et al., 2012, for a discussion), recent experimental research affords key evidence that enthymematic validity (Reimer, 2003; Reimer et al., 2016) and the validity and distinctiveness of cues can serve as criteria of the actual quality of an argument (Reimer et al., 2012; Russell et al., 2014). The actual quality of an argument refers to the theoretically derived features of an argument that define its quality independent of receivers’ evaluation (O’Keefe, 2003; Reimer et al., 2012). The perceived quality of an argument, conversely, refers to receivers’ perception of argument quality (O’Keefe, 2003; Reimer et al., 2012). With the goal of contributing to this research tradition, the current project introduced two novel criteria of argument quality, namely attribute degree centrality and attribute tie strength. Both criteria are thought to provide theory-based criteria of argument quality that are related to the actual quality of arguments. Though an empirical question, we suggest both attribute dimensions may influence decision-making and persuasion outcomes under specific task conditions through the mediating mechanism of spreading activation in semantic memory. Conclusions and Future directions The most critical future direction for research involves experimentally testing whether, and to what extent, highly central attributes influence decision-making and persuasion outcomes. We predict that by augmenting the mind’s ability to make inferences and choices under uncertainty, highly central attributes will indeed be more persuasive. As a critical step in accomplishing this goal, the present research examined semantic networks in the environment and across cognitive representations. Future research should aim to choose attributes to embed in arguments that are high (and low) in weighted degree centrality across both the environment and cognitive representations to test whether an attribute’s number and strength of connections provide external argument quality criteria and influence the persuasiveness of communicators. In our research, we used headphones as a real-world example since it is an everyday life decision people make but one not typically accompanied by extremes in terms of preferences or knowledge. Selecting examples using similar choice criteria with respect to familiarity, preference, and knowledge, future studies should seek to formally replicate the present research. In addition to formal replication, future directions for research may also focus on examining semantic network structure in the external world and cognitive representations pertaining to different domains, for example regarding attributes of food items, political candidates, or financial options. This research has emphasized pragmatic arguments arguing for choice of a focal object (e.g., Purchase the Loneg JBT-1380) that has positive attributes (e.g., because it has Sound Isolating Technology). Future research should experimentally examine variations in message design corresponding to differences in the way choice alternatives and arguments are expressed and combined. For example, there are eight ways to construct a message with a specific choice alternative (e.g., for or against choice of a focal object) and a particular cue or attribute (e.g., a highly or lowly central, and a positive or negative attribute). The current conceptualization focused on positive arguments that speak in favor of a focal object. Theoretically, we suggest arguments for a focal object utilizing a negative attribute strongly connected to many negative attributes would operate as a contra argument decreasing the choice of the focal object. Future research should test these predictions and other similar hypotheses related to different message variations. In the current research, attribute degree centrality and tie strength were operationalized in terms of external features of an object focusing on pragmatic argumentation. However, the use of semantic network models and theory within the PPT context are not limited to pragmatic argumentations that focus on the consequences of decisions. PPT specifies four categories of cues, namely object, speaker, listener, and context cues that can be used as arguments in different judgment and decision contexts; tasks including multiple alternative choice, estimation, and classification (see Reimer et al., 2012, for a discussion). The acquisition of semantic network structure and spreading activation processes are not restricted to one category of cues and their applicability is not limited to one type of decision task. Given its theoretical generality and flexibility, the current project may be profitably combined with argumentation research which identifies common types of argument (e.g., argument from authority) and proposes cues relevant to the quality of each argument type (e.g., evidence of professional credentials) (see Walton, Reed, & Macagno, 2008, for an overview). Comparable to the Bayesian approach (see Hahn & Hornikx, 2016, for a discussion), the present approach provides a useful means for determining what type-relevant cues are high in quality and potentially persuasive in an argumentation scenario. References Begg, I., Duft, S., Lalonde, P., Melnick, R., & Sanvito, J. ( 1989). Memory predictions are based on ease of processing. Journal of Memory and Language , 28( 5), 610– 632. doi:10.1016/0749-596X(89)90016-8 Google Scholar CrossRef Search ADS   Boller, G. W., Sway, J. L., & Munch, J. M. ( 1990). Conceptualizing argument quality via argument structure. Advances in Consumer Research , 17( 1), 321– 328. http://acrwebsite.org/volumes/9831/volumes/v17/NA-17 Brunswik, E. ( 1935). Experimentelle psychologie in demonstrationen . Wien: J. Springer. http://dx.doi.org/10.1007/978-3-7091-9901-5 Google Scholar CrossRef Search ADS   Brunswik, E. ( 1944). Distal focussing of perception: Size-constancy in a representative sample of situations. Psychological Monographs: General and Applied , 56( 1), 1– 49. http://dx.doi.org/10.1037/h0093505 Google Scholar CrossRef Search ADS   Brunswik, E. ( 1955). Representative design and probabilistic theory in a functional psychology. Psychological Review , 62( 3), 193– 217. http://dx.doi.org/10.1037/h0047470 Google Scholar CrossRef Search ADS   Chang, T. M. ( 1986). Semantic memory: Facts and models. Psychological Bulletin , 99( 2), 199– 220. http://dx.doi.org/10.1037/0033-2909.99.2.199 Google Scholar CrossRef Search ADS   Collins, A. M., & Loftus, E. F. ( 1975). A spreading-activation theory of semantic processing. Psychological Review , 82( 6), 407– 428. http://dx.doi.org/10.1037/0033-295X.82.6.407 Google Scholar CrossRef Search ADS   Collins, A. M., & Quillian, M. R. ( 1969). Retrieval time from semantic memory. Journal of Verbal Learning and Verbal Behavior , 8( 2), 240– 247. doi:10.1016/S0022-5371(69)80069-1 Google Scholar CrossRef Search ADS   Corey, D. M., Dunlap, W. P., & Burke, M. J. ( 1998). Averaging correlations: Expected values and bias in combined Pearson rs and Fisher’s z transformations. The Journal of General Psychology , 125( 3), 245– 261. doi:10.1080/00221309809595548 Google Scholar CrossRef Search ADS   Cowley, E., & Mitchell, A. A. ( 2003). The moderating effect of product knowledge on the learning and organization of product information. Journal of Consumer Research , 30( 3), 443– 454. http://dx.doi.org/10.1086/378620 Google Scholar CrossRef Search ADS   Fink, E. L., High, A. C., & Smith, R. A. ( 2015). Compliance dynamics within a friendship network II: Structural positions used to garner social support. Human Communication Research , 41( 1), 21– 54. doi:10.1111/hcre.12038 Google Scholar CrossRef Search ADS   Fink, E. L., Monahan, J. L., & Kaplowitz, S. A. ( 1989). A spatial model of the mere exposure effect. Communication Research , 16( 6), 746– 769. doi:10.1177/009365089016006002 Google Scholar CrossRef Search ADS   Freeman, L. C. ( 1979). Centrality in social networks conceptual clarification. Social Networks , 1( 3), 215– 239. doi:10.1016/0378-8733(78)90021-7 Google Scholar CrossRef Search ADS   Gigerenzer, G., Czerlinski, J., & Martignon, L. ( 1999). How good are fast and frugal heuristics? In J. Shanteu, B. A. Mellers, & D. A. Schum (Eds.), Decision science and technology  (pp. 81– 103). New York: US Springer. doi:10.1007/978-1-4615-5089-1_6 Google Scholar CrossRef Search ADS   Gigerenzer, G., & Goldstein, D. G. ( 1996). Reasoning the fast and frugal way: Models of bounded rationality. Psychological Review , 103( 4), 650– 669. http://dx.doi.org/10.1037/0033-295X.103.4.650 Google Scholar CrossRef Search ADS   Gigerenzer, G., Hoffrage, U., & Kleinbölting, H. ( 1991). Probabilistic mental models: A Brunswikian theory of confidence. Psychological Review , 98( 4), 506– 528. http://dx.doi.org/10.1037/0033-295X.98.4.506 Google Scholar CrossRef Search ADS   Gigerenzer, G., Todd, P. M. & the ABC Research Group. ( 1999). Simple heuristics that make us smart . New York: Oxford University Press. Glucksberg, S., & McCloskey, M. ( 1981). Decisions about ignorance: Knowing that you don’t know. Journal of Experimental Psychology: Human Learning and Memory , 7( 5), 311– 325. http://dx.doi.org/10.1037/0278-7393.7.5.311 Google Scholar CrossRef Search ADS   Graesser, A. C., & Hemphill, D. ( 1991). Question answering in the context of scientific mechanisms. Journal of Memory and Language , 30( 2), 186– 209. doi:10.1016/0749-596X(91)90003-3 Google Scholar CrossRef Search ADS   Grebitus, C., & Bruhn, M. ( 2006). Consumers’ demand for pork quality: Applying semantic network analysis. Card Working Paper 06-WP 423. http://lib.dr.iastate.edu/card_workingpapers/456 Hahn, U., & Hornikx, J. ( 2016). A normative framework for argument quality: Argumentation schemes with a Bayesian foundation. Synthese , 193( 6), 1833– 1873. doi:10.1007/s11229-015-0815-0 Google Scholar CrossRef Search ADS   Hahn, U., & Oaksford, M. ( 2006). A Bayesian approach to informal argument fallacies. Synthese , 152( 2), 207– 236. doi:10.1007/s11229-005-5233-2 Google Scholar CrossRef Search ADS   Hahn, U., & Oaksford, M. ( 2007). The rationality of informal argumentation: A Bayesian approach to reasoning fallacies. Psychological Review , 114( 3), 704– 732. http://dx.doi.org/10.1037/0033-295X.114.3.704 Google Scholar CrossRef Search ADS   Hample, D., & Dallinger, J. M. ( 1990). Arguers as editors. Argumentation , 4( 2), 153– 169. doi:10.1007/BF00175420 Google Scholar CrossRef Search ADS   Harris, A. J., Hsu, A. S., & Madsen, J. K. ( 2012). Because Hitler did it! Quantitative tests of Bayesian argumentation using ad hominem. Thinking & Reasoning , 18( 3), 311– 343. doi:10.1080/13546783.2012.670753 Google Scholar CrossRef Search ADS   Henley, N. M., Horsfall, R. B., & De Soto, C. B. ( 1969). Goodness of figure and social structure. Psychological Review , 76( 2), 194– 204. http://dx.doi.org/10.1037/h0027358 Google Scholar CrossRef Search ADS   Hertwig, R., Herzog, S. M., Schooler, L. J., & Reimer, T. ( 2008). Fluency heuristic: A model of how the mind exploits a by-product of information retrieval. Journal of Experimental Psychology: Learning, Memory, and Cognition , 34( 5), 1191– 1206. http://dx.doi.org/10.1037/a0013025 Google Scholar CrossRef Search ADS   Hoeken, H., Šorm, E., & Schellens, P. J. ( 2014). Arguing about the likelihood of consequences: Laypeople’s criteria to distinguish strong arguments from weak ones. Thinking & Reasoning , 20( 1), 77– 98. doi:10.1080/13546783.2013.807303 Google Scholar CrossRef Search ADS   Hoeken, H., Timmers, R., & Schellens, P. J. ( 2012). Arguing about desirable consequences: What constitutes a convincing argument? Thinking & Reasoning , 18( 3), 394– 416. doi:10.1080/13546783.2012.669986 Google Scholar CrossRef Search ADS   Hoffrage, U., & Reimer, T. ( 2004). Models of bounded rationality: The approach of fast and frugal heuristics. Management Revue , 15( 4), 437– 459. Hornikx, J., & Hoeken, H. ( 2007). Cultural differences in the persuasiveness of evidence types and evidence quality. Communication Monographs , 74( 4), 443– 463. doi:10.1080/03637750701716578 Google Scholar CrossRef Search ADS   Jonassen, D. H., Beissner, K., & Yacci, M. ( 1993). Structural knowledge: Techniques for representing, conveying, and acquiring structural knowledge . New Jersey: Hillsdale. Katsikopoulos, K. V., Schooler, L. J., & Hertwig, R. ( 2010). The robust beauty of ordinary information. Psychological Review , 117( 4), 1259– 1266. doi:10.1037/a0020418 Google Scholar CrossRef Search ADS   Kelley, C. M., & Lindsay, D. S. ( 1993). Remembering mistaken for knowing: Ease of retrieval as a basis for confidence in answers to general knowledge questions. Journal of Memory and Language , 32( 1), 1– 24. doi:10.1006/jmla.1993.1001 Google Scholar CrossRef Search ADS   Koriat, A., Lichtenstein, S., & Fischhoff, B. ( 1980). Reasons for confidence. Journal of Experimental Psychology: Human Learning and Memory , 6( 2), 107– 118. http://dx.doi.org/10.1037/0278-7393.6.2.107 Google Scholar CrossRef Search ADS   Kruglanski, A. W., & Thompson, E. P. ( 1999). Persuasion by a single route: A view from the unimodel. Psychological Inquiry , 10( 2), 83– 109. doi:10.1207/S15327965PL100201 Google Scholar CrossRef Search ADS   Lee, S., & Monge, P. ( 2011). The coevolution of multiplex communication networks in organizational communities. Journal of Communication , 61( 4), 758– 779. doi:10.1111/j.1460-2466.2011.01566.x Google Scholar CrossRef Search ADS   Loftus, E. F. ( 1973). Category dominance, instance dominance, and categorization time. Journal of Experimental Psychology , 97( 1), 70– 74. http://dx.doi.org/10.1037/h0033782 Google Scholar CrossRef Search ADS   Mercier, H., & Sperber, D. ( 2011). Why do humans reason? Arguments for an argumentative theory. Behavioral and Brain Sciences , 34( 2), 57– 74. doi:10.1017/S0140525X10000968 Google Scholar CrossRef Search ADS   Monge, P. R., & Contractor, N. S. ( 2003). Theories of communication networks . New York: Oxford University Press. Morris, C. C. ( 1990). Retrieval processes underlying confidence in comprehension judgments. Journal of Experimental Psychology: Learning, Memory, and Cognition , 16( 2), 223– 232. http://dx.doi.org/10.1037/0278-7393.16.2.223 Google Scholar CrossRef Search ADS   Nelson, T. O., & Narens, L. ( 1990). Metamemory: A theoretical framework and new findings. In G. H. Bower, (Ed.), The psychology of learning and motivation: Advances in research and theory  (Vol. 26, pp. 125– 169). San Diego, CA: Academic Press. O’Keefe, D. J. ( 2003). Message properties, mediating states, and manipulation checks: Claims, evidence, and data analysis in experimental persuasive message effects research. Communication Theory , 13( 3), 251– 274. doi:10.1111/j.1468-2885.2003.tb00292.x Google Scholar CrossRef Search ADS   O’Keefe, D. J. ( 2013). The relative persuasiveness of different forms of arguments-from-consequences: A review and integration. Annals of the International Communication Association , 36( 1), 109– 135. http://dx.doi.org/10.1080/23808985.2013.11679128 Google Scholar CrossRef Search ADS   O’Keefe, D. J., & Jackson, S. ( 1995). Argument quality and persuasive effects: A review of current approaches. In S. Jackson (Ed.), Argumentation and values: Proceedings of the ninth Alta conference on argumentation  (pp. 88– 92). Annandale, VA: Speech Communication Association. Perelman, C. ( 1959). Pragmatic arguments. Philosophy , 34( 128), 18– 27. https://doi.org/10.1017/S0031819100029697 Google Scholar CrossRef Search ADS   Peter, J. P., Olson, J. C., & Grunert, K. G. ( 1999). Consumer behavior and marketing strategy  (pp. 122– 123). London: McGraw-Hill. Reimer, T. ( 1997). Präsupponieren und Argumentieren: Präsuppositionen als zentrale Hinweisreize bei einer elaborierten Verarbeitung von Argumenten. [Presupposing and argumentation: Presuppositions as central cues under high argument elaboration.] Sprache und Kognition , 2, 61– 77. Reimer, T. ( 2003). Direkte und indirekte Effekte der Argumentqualität: Der Einfluss der Argumentstärke auf die wahrgenommene Expertise eines Kommunikators. [Direct and indirect effects of argument quality: The impact of argument strength on the perceived expertise of a communicator.] Zeitschrift für Sozialpsychologie , 34( 4), 243– 255. http://dx.doi.org/10.1024/0044-3514.34.4.243 Google Scholar CrossRef Search ADS   Reimer, T., Hertwig, R., & Sipek, S. ( 2012). Probabilistic persuasion: A Brunswikian theory of argumentation. In R. Hertwig, U. Hoffrage, & the ABC Research Group (Eds.), Simple heuristics in a social world  (pp. 33– 55). New York: Oxford University Press. https://doi.org/10.1093/acprof:oso/9780195388435.003.0004 Reimer, T., Raab, M., & Russell, T. ( 2017). The message-primacy hypothesis: Effects of argument quality on the evaluation of proponents. Manuscript under review. Reimer, T., & Rieskamp, J. ( 2007). Fast and frugal heuristics. In R. F. Baumeister & K. D. Vohs (Eds.), Encyclopedia of social psychology  (pp. 346– 348). Thousand Oaks, CA: Sage. http://dx.doi.org/10.4135/9781412956253.n215 Russell, T., Reimer, T., Hertwig, R. ( 2014). The probabilistic persuasion theory: Exploring the effects of argument quality and communication modality on consumer decision making. Presented at the 100th Annual Convention of the National Communication Association, Chicago, IL. Smith, E. E., & Kosslyn, S. M. ( 2007). Cognitive psychology: Mind and brain . Upper Saddle River, NJ: Pearson Prentice Hall. Stiff, J. B. ( 1986). Cognitive processing of persuasive message cues: A meta‐analytic review of the effects of supporting information on attitudes. Communications Monographs , 53( 1), 75– 89. doi:10.1080/03637758609376128 Google Scholar CrossRef Search ADS   Todd, P. M., & Gigerenzer, G. ( 2012). What is ecological rationality? In P. M. Todd, G. Gigerenzer, & the ABC Research Group (Eds.), Ecological rationality: Intelligence in the world  (pp. 3– 33). New York: Oxford University Press. Google Scholar CrossRef Search ADS   Tolman, E. C., & Brunswik, E. ( 1935). The organism and the causal texture of the environment. Psychological Review , 42( 1), 43– 77. http://dx.doi.org/10.1037/h0062156 Google Scholar CrossRef Search ADS   Trouche, E., Johansson, P., Hall, L., & Mercier, H. ( 2015). The selective laziness of reasoning. Cognitive Science , 42( 1), 2122– 2136. doi:10.1111/cogs.12303 Tulving, E. ( 1972). Episodic and semantic memory. In E. Tulving & W. Donaldson (Eds.), Organization of memory , (pp. 381– 402). Cambridge: Academic Press. Walton, D. ( 2005). Justification for argumentation schemes. Australasian Journal of Logic , 3, 1– 13. Walton, D., Reed, C., & Macagno, F. ( 2008). Argumentation schemes . Cambridge: Cambridge University Press. Google Scholar CrossRef Search ADS   Wasserman, S., & Faust, K. ( 1994). Social network analysis: Methods and applications . Cambridge: Cambridge University Press. Google Scholar CrossRef Search ADS   Whittlesea, B. W. ( 1993). Illusions of familiarity. Journal of Experimental Psychology: Learning, Memory, and Cognition , 19( 6), 1235– 1253. http://dx.doi.org/10.1037/0278-7393.19.6.1235 Google Scholar CrossRef Search ADS   Winkielman, P., & Cacioppo, J. T. ( 2001). Mind at ease puts a smile on the face: Psychophysiological evidence that processing facilitation elicits positive affect. Journal of Personality and Social Psychology , 81( 6), 989– 1000. http://dx.doi.org/10.1037/0022-3514.81.6.989 Google Scholar CrossRef Search ADS   Woelfel, J., & Fink, E. L. ( 1980). The measurement of communication processes: Galileo theory and method . New York: Academic Press. Woelfel, J., Gillham, J., Cody, M. J., & Holmes, R. A. ( 1980). Basic premises of multidimensional attitude change theory: An experimental analysis. Human Communication Research , 6( 2), 153– 167. doi:10.1111/j.1468-2958.1980.tb00135.x Google Scholar CrossRef Search ADS   Zhao, X., Strasser, A., Cappella, J. N., Lerman, C., & Fishbein, M. ( 2011). A measure of perceived argument strength: Reliability and validity. Communication Methods and Measures , 5( 1), 48– 75. doi:10.1080/19312458.2010.547822 Google Scholar CrossRef Search ADS   © The Author(s) 2018. Published by Oxford University Press on behalf of the International Communication Association. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)
TI - Using Semantic Networks to Define the Quality of Arguments
JF - Communication Theory
DO - 10.1093/ct/qty003
DA - 2018-02-01
UR - https://www.deepdyve.com/lp/oxford-university-press/using-semantic-networks-to-define-the-quality-of-arguments-pEIsXTb0KA
SP - 46
EP - 68
VL - 28
IS - 1
DP - DeepDyve
ER -