- Email: [email protected]
Understanding reading as a form of language-use: A language game hypothesis
New Ideas in Psychology xxx (2015) 1e8
Contents lists available at ScienceDirect
New Ideas in Psychology journal homepage: www.elsevier.com/locate/newideapsych
Understanding reading as a form of language-use: A language game hypothesis Sebastian Wallot* Interacting Minds Centre, Department of Culture and Society, Aarhus University, Denmark
a r t i c l e i n f o
a b s t r a c t
Article history: Received 15 March 2015 Received in revised form 17 July 2015 Accepted 20 July 2015 Available online xxx
Reading research and research on conversation have followed different paths: While the research program for reading committed itself to a relatively static view of language, where objective text properties serve to elicit specific effects on cognition and behavior of a reader, research on conversation has embraced a language-use perspective, where language is primarily seen as a dynamic, context dependent process. In this essay I contrast these two perspectives, and argue that in order to reach a unified understanding of natural language e be it reading, talking, or conversing e one needs to adopt a languageuse perspective. Furthermore, I describe how reading can be seen as a form of language-use, and how the current landscape of research on reading can be re-interpreted in terms of a dynamic, context-sensitive perspective on language. In particular, I propose that the concept of ‘language games’ serves as a good starting point to conceive reading as a form of language-use, describe how one can derive first concrete hypotheses by re-interpreting reading in terms of language games, and show how they can be readily operationalized using tools from dynamic systems analysis. © 2015 Elsevier Ltd. All rights reserved.
Keywords: Reading Conversation Language-use Dynamic systems Language games
1. Introduction Reading is a cultural-cognitive skill that most people exercise on a daily basis. From a psychological perspective, reading is interesting because of this cognitive-cultural aspect: As a cultural skill, reading is of practical relevance for education, economy, and recreation. As a cognitive skill, reading touches many aspects of human cognition and interaction. Reading is a form of natural languageuse, just as talking, listening, conversing. It is communication, understanding, thinking, and perception. Reading has evolved for communicative purposes, to pass along messages, but also to extend the mind into the environment, for example to enhance memory by writing notes or to reflect upon one's own thoughts by writing them down. Obviously, reading mandates perception, as text is a property of the environment, and needs to be seen in order to be utilized. However, the perceptual side of reading has received much more scientific attention than any of its other aspects. The main puzzle that psychologists sought to solve for reading is, how visual contrast gradients in the environment can be identified as words/
* Interacting Minds Centre, Aarhus University Jens Chr. Skous Vej 4, Building 1.483, 3rd floor, 8000 Aarhus C, Denmark. E-mail address: [email protected].
language by readers. Language can be written/printed in many diverse and unique ways, and success with extracting language from an environmental sources has been limited to clean print, or isolated letters (Plamondon & Srihari, 2000). Hence, research has taken a half-step back from the straight question of how visual features are recognized by a reader, and settled instead on linguistic features that help a reader to recognize a particular word of written language. Such features, called lexical variables, are the centerpiece of the well-developed contemporary theories and models of reading, implemented in computational models of word recognition and models of eye-movement control during reading (e.g., Grainger & Jacobs, 1996; Reichle, Rayner, & Pollatsek, 2003). Lexical variables are quantifiable aspects of a word, usually based on corpus-linguistic analyses. One of the most prominent lexical variables is called word frequency. In order to quantify the frequency of a word, one counts how often that word appears in a representative corpus of texts. The relative number of appearances of that word in the corpus is then an estimator of its frequency. Quantitative measures of word frequency can be correlated with quantitative measures of the reading process (e.g., reaction times, fixation durations …), and further refinements can be made, for example the prominent logarithmic transformation of the frequency values, or theoretical parameterization of a regression model that converts frequencies into gains in terms of reading
http://dx.doi.org/10.1016/j.newideapsych.2015.07.006 0732-118X/© 2015 Elsevier Ltd. All rights reserved.
Please cite this article in press as: Wallot, S., Understanding reading as a form of language-use: A language game hypothesis, New Ideas in Psychology (2015), http://dx.doi.org/10.1016/j.newideapsych.2015.07.006
2
S. Wallot / New Ideas in Psychology xxx (2015) 1e8
speed in milliseconds. The basic idea is that words exhibit stable linguistic characteristics that are captured by lexical variables, which in turn allow a reader to reliably identify words, the assumed building blocks of written language. A second strand of research that is also relatively prominent in reading research is concerned with comprehension processes of the reader (e.g., Graesser & McNamara, 2011). Interestingly, the research that focuses on the perceptual front-end of reading and the research that focuses on later-stage comprehension processes is relatively disjoint, with the former focusing heavily on process measures of reading, and the latter focusing on outcome measures of reading (but there are ongoing attempts to bring results from comprehension research into process measures of reading e Reichle, Warren, & McConnell, 2009). Conceptually, however, comprehension research is well aligned at its meta-theoretical core with the perceptual/process research on reading. Just as the perceptual/process research on reading assumes that a particular word exhibits particular features that lead to particular effects on the side of the reader, comprehension focused research assumes that comprehension of a text implies comprehension of something particular. That is, that there are stable features, for example on the sentence level, that act either as a comprehension-marker of that sentence or that connect two adjacent sentences (e.g., situation model dimensions), and that those features are reliably used by readers to comprehend a particular message (cf. Graesser & McNamara, 2011). To summarize, reading research e on the level of perception and on the level of comprehension e has thrived on the assumption that the reading process is specific. It is specific in the sense that there are specific aspects of texts (e.g., certain lexical characteristics of words) that map onto specific cognitive processes (e.g., word identification processes) and specific cognitive contents (e.g., the meanings of words as prescribed in their representation in the mental lexicon). In order to be specific in this way, there must be a stable relationship between aspects of the text and responses of a reader with regard to the text. One can only establish a link between, say, a word property and a particular mental content that in turn triggers a measureable response (e.g., a reaction time, a fixation duration, an event-related potential) if both sides on this mapping relationship remain stable. Hence, lexical variables such as word frequency are only useful as universtal building blocks for theories of reading if the frequency of a word is relatively constant, and if the meaning of a word remains more or less the same across different reading situations. If, for example, the meaning of a word can change radically or instantaneously in different reading situations, then word frequency would not be of any help to link the visual impression of a word to its meaning. The frequency of a word would not be specific any more with regard to the particular mental representation it supposedly helps to index. The guiding assumption of specific relations between text properties and cognitive processes/contents has let to a fruitful e maybe even the most fruitful e research program in experimental psychology, and has generated practical results that have spilled over in the world of every-day readers, such as the automated readability metrics that are now part of many word processing programs (e.g., the Flesh-Kincaid index), or the design of speedreading software (e.g., implementation of the optimal viewing position effect in RSVP reading software). At the same time, however, research on reading has lived a somewhat exclusive life in its own traditional niche, apart from other natural language activities, especially conversation. Some researchers have tried to establish a conceptual link between conversation/spoken language and reading (Chafe & Tannen, 1987), but empirical research on the topic is virtually absent. The main reason why it seems so difficult to reconcile reading with
conversation seems to be the diverging conception in these two research traditions with regard to what language is, and how it inherently works. Compared to research on reading, which has focused on perception and the mapping of stable word features to cognition, research on conversation, communication and social interaction stands on the other end of the spectrum. Research on reading might have taken the view of reading as a strongly componential process as a result of mainly dealing with pieces of written language, which seem to convey a static picture of language because that is how texts make language appear, exhibiting its own version of the what has been called the written language bias (Linell, 2005) in linguistics. In contrast, research on conversation has adopted a perspective with different emphasis, namely how its bearers use language. Note that this ontological divergence in the conception of language should strike one as odd, as a naïve perspective on language development would suggest that the two are closely related: After all, reading always develops on the language capabilities that are based on conversation and social interaction in infancy and childhood. So intuitively, these two domains of language should not work according to separate principles, but share a common core. Compared to the componential/perceptual stance adopted by reading research, research on conversation has adopted a usagebased stance, which offers a strongly contextualized perspective on language e i.e., that it is not a set of (perceptual mapping) rules that structure language, but that language is structured by functional aspects that differ from context to context in which it is used. Among other things, this reserves a fundamental role for the intentions of participants in a communication, for the social context and purpose of the communicative situation, as well as for the dynamics of conversation e that is, that communication is not built out of static components, but that the development of a communicative process is consequential for the outcome of that commun, 2014). nication (Fusaroli, Ra˛ czaszek-Leonardi, & Tyle Admittedly, all these considerations seem to be fine, if not inevitable, for such a thing as two people standing in the same room, being able to talk to each other, see and hear each other, and interact with each other. However, from the surface, reading seems quite different, where a single person sits or stands still, scanning the environment for patterns. So how does all of this pertain to reading at all? And why should one even bother? 2. Reading is complex, not just complicated Perhaps the reason why one regards the phenomenon 'conversation' as being appropriately described by an account that allows central roles for intentions, context and dynamics and shies away from an all too strict reductionism is, because conversation is easily seen as a complex phenomenon: Research on language in social interaction (Clark, 1996), as well as common everyday experience suggest that an abundance of interdependent factors play a role for conversation, and that conversation can either go wrong or succeed in unexpected ways, no matter how clear the goals of the interlocutors or their communicative ability, or how seemingly simple the topic. For example in a perceptual identification task, two participants see each the same six striped patterns on a computer monitor, one being slightly different from the other five. They are to decide which of the six is the odd one and individually submit their decision to the experimenter. If the two do not agree, then they are allowed to take a break and discuss which choice they should go with. Interestingly, each pair develops its very own vocabulary to deal with this kind of task (Fusaroli et al., 2012). For example, two participants developed a cheese-confidence-scale, where each participant told the other how confident they were that their choice was correct in ascending orders of cheese-stinkyness (“Cheese. It's
Please cite this article in press as: Wallot, S., Understanding reading as a form of language-use: A language game hypothesis, New Ideas in Psychology (2015), http://dx.doi.org/10.1016/j.newideapsych.2015.07.006
S. Wallot / New Ideas in Psychology xxx (2015) 1e8
cheese. It's stinky cheese. Stilton!” e Riccardo Fusaroli, personal communication). This odd and humorous example should simply serve to illustrate that successful communication during conversation can take an abundance of forms, and it is hard to pin them down to a specific corpus of words that would, for example, delimit good from bad interactions. It is rather the functionality that is driving conversational language choice, not a stereotyped lexicon. However, when one thinks about reading, it becomes clear that there are similarities between reading and conversation beyond the emphasis of the perceptual level: Just as communication in a successful conversation depends on mutual consideration of the intentions of interlocutors, reading depends on the intentions of the reader. Readers' intentions have consequences for how reading unfolds over time, for reading behavior, emotional experience, and memory (Hunt & Vipond, 1991). Furthermore, reading also depends on the (assumed and real) intentions of the author, who becomes the equivalent of an interlocutor in a real conversation (Davidson, 1993). Again, however, the role of the author is not static: As the text unfolds, the author's intentions need to be revised, such as when irony is used, or the reader is ‘led down the wrong path’. Just as in real conversations, where interlocutors go back and forth to establish common ground or pause for reflection (Fusaroli et al., 2012), research on reading has observed feedback loops during reading, such as re-reading previous text passages (Rayner, Chace, Slattery, & Ashby, 2006), a sort of substitute for online conversational exchange. Hence, seen this way, there is a substantial overlap between how conversation and reading work, and as one approaches naturalistic texts e as opposed to, say, random word lists e these factors, which are hard to pinpoint to objective text features, begin to matter more and more. In addition to this more phenomenological side, reading has also proved to be complex with regard to the strong commitment towards its anchoring on linguistic properties, such as lexical variables, syntax, and text-structure based measures of coherence and semantics. If one considers the simplest reading tasks, such as lexical decision or word naming, where lexical word features do play a substantial role, the complexity of reading reveals itself through nonlinear correlations between the different lexical features that are thought to drive reading, as well as through the effects of these features on reading performance: In general, the majority of lexical features (such as word frequency, word length, co-occurrence, neighborhood frequency etc.) are highly correlated amongst each other. Moreover, the correlations among these features are not linear with regard to reading performance: For example, the effects of word frequency depend on the semantic context within which a word is presented (Van Petten & Kutas, 1991), or orthographic neighborhood frequency (Grainger, 1990). The heterogeneity of findings in reading is a problem because this makes it difficult to say which role a particular factor that is investigated in reading research actually plays for the reading process e and this problem is not limited to the level of word reading. Prominent sentence-reading tasks reveal that the role of word properties is dependent on syntactic complexity (Keller, Carpenter, & Just, 2001). Moreover, eye-movement studies during reading have generally called into question in how far individual words can be treated as separable entities: When a reader can see more than one word at a time, words at the periphery of vision are already being pre-processed by the reader (Hohenstein & Kliegl, 2014), and this parafoveal preview influences the reading of the word that is currently being fixated by the reader, generating facilitative or inhibitory effects for the reading process depending on the relation between these two words. As we scale up once more from sentence to connected text reading, lexical features loose their constitutive role for the reading process all together (Wallot, Hollis, & van Rooij, 2013). Similarly,
3
text coherence metrics that were developed to characterize text reading fail to yield many of their predicted effects on a reader's performance, if one observes reading of long connected text that spans more than a few paragraphs (McNerney, Goodwin, & Radvansky, 2011). One can say that it is in fact not even clear what drives the reading process of naturalistics text, as it seems that many of the relatively simple relations between objective text features and reader performance that are observed in simpler reading tasks fail to reveal themselves in naturalistic reading. It is almost as if naturalistic reading would be an entirely different phenomenon, at least something else than a concatenation of words or sentences. Finally, the different kinds of reading tasks that have been used to measure reading in psychological laboratories add another layer of complexity on top of what different kinds of text materials provide: Reading word lists silently versus aloud already changes the way that reading performance and text properties relate (Grainger, 1990), and such interactions between task performance and context seem to be the rule, rather than the exception (Van Orden, Pennington, & Stone, 2001). The above examples are meant to illustrate that reading is not merely complicated, but complex. If reading were only complicated, then reading research is on the right track and by describing a great number of relevant factors, a relatively simple and coherent theory of reading would stand at the end of those efforts. However, such a theory is nowhere in sight (Rayner & Reichle, 2010) and it seems that reading is more than just complicated. The factors that are investigated in reading are intricately intertwined with each other, and change their role and relation as linguistic, but also environmental aspects of a reading situation are changed. Hence, reading is not just complicated, it is complex, and with the examples above I wanted to illustrate how some of the assumptions of a language-use perspective readily also pertain to reading research that has largely been conducted under different premises. In particular, it seems that the presence of an author that provides for an intentional background of a reading task changes how reading unfolds, something which is (largely) absent in the prominent isolated word and sentence reading tasks e a random word list is very hard to read as an intended message. Moreover, non-linguistic aspects seem to highly influence the reading process as well, such as how a text is displayed (one word at a time, one sentence at a time, one page at a time e Wallot & Van Orden, 2011), what tasks and measurement setups are utilized (Grainger, 1990), and what situation a participant is asked to perform the reading in (Hunt & Vipond, 1991). Here, one might see traces of what have been thought to be genuine conversational properties e intention, situation, and non-linguistic context e in reading. That is, effects of things that are themselves not textual entities, but co-determine how language is perceived and understood by a reader. The question is, what would be an adequate framework for conceiving the situation of a text and a reader in terms of languageuse? Such a framework should be connectable to conversation, but also provide initial hypotheses and an adequate measurementframework for empirical investigations of the reading process. After all, prominent measures that capture aspects of language-use in studies on conversation, such as coupling of movements (Richardson, Dale, & Kirkham, 2007) or terminology (Fusaroli et al., 2012) between participants, cannot be obtained from a reading situation. In the next section, I will outline such a framework, based on the concept of language games, originally stipulated by Ludwig Wittgenstein in the ‘Philosophical Investigations’, and show how analytic tools from the toolbox of dynamic systems can be used to operationalize these concepts.
Please cite this article in press as: Wallot, S., Understanding reading as a form of language-use: A language game hypothesis, New Ideas in Psychology (2015), http://dx.doi.org/10.1016/j.newideapsych.2015.07.006
4
S. Wallot / New Ideas in Psychology xxx (2015) 1e8
3. Conceiving reading as language use As I have proposed elsewhere (Wallot, 2014), the concepts of form-of-life and language games (Wittgenstein, 1953/2010) might serve as a theoretical anchor for a unifying, usage-based perspective on conversation and reading, with implications for how research on reading as language-use can be done. Both of these concepts were most likely formulated with conversational situations in mind, not with reading. However, as I will explain in the following, these two concepts lend themselves well to an application of the reading situation. In the ‘Philosophical Investigations’, linguistic structure is described as a variant phenomenon, and contrary to the dominant linguistic perspective, objective structures in language (such as syntax) are not primarily seen as building blocks of language, its production, or comprehension, but as emergent phenomena. This follows from the two related concepts, form-of-life and language games: Form-of-life describes the idea that language is always used within a set of (non-linguistic) constraints that shape and constitute the contexts within which people use language. These constraints can span historical influences, geographical and material aspects of the current environment in which a conversation takes place, as well as the social norms and personal characteristics and relationships of the conversants. Form-of-life supplies the repertoire of potentially meaningful language-uses. Within form-of-life, different language games can be played. Language games can be seen as a concrete realization from the repertoire supplied by form-of-life. They are an actual communicative exchange that can be observed. Wittgenstein (1953/2010) chose the term game, because a concrete instance of a conversation looks like a rule-abiding activity that is functional towards keeping the communicative exchange going. However, just as there are many different games with many different rules, the goals and nature of communicative situations are very diverse. Moreover, as the nature of a communicative situation changes, so do the rules according to which that communication unfolds. Hence, the rules that are observed in the language between two interlocutors are not a reflection of the principal architecture of language, but they are themselves emergent, being realized within the constraints of form-of-life, the actual situation in which that communication takes place. Let us consider the application of form-of-life and language games to reading, and preliminary hypotheses for reading research that can be derived from these concepts: First of all, the language game perspective shares a sentiment regarding the effects of language on a person's (linguistic) behavior with classical reading research, namely that language-use e reading e is a rule-abiding activity. However, for the majority of reading research, this has translated into a stricter assumption, namely that reading can be defined as a specific reaction of the reader to specific aspects of objective text structure. That is, the structure observed in reading behavior is to be traced back to objective text structures (lexical variables, syntactical features …), and an understanding of these structures would mean to understand the universal rules by which reading behavior abides. Moreover, the relation between text and reader can be described by specific rules (i.e., the higher the frequency of a word, the shorter the fixation duration of that word), which would allow to predict a reader's behavior across any reading situation. In the conception of language games, rules play a different role: Reading is also seen as a structured activity, and in a particular reading situation (e.g., reading random word lists and responding to each word with a button press), one can observe and formulate particular rules according to which the observed reading process behaves. However, these rules are not universal, and are themselves
a result of the reading situation. As one changes the reading situation, that is, changes the language game, partially or completely different rules might be observed that guide a reader's behavior. This seems also consonant with the observation that any particular rule that has been formulated by reading researchers to govern the reading process has been observed to be variant, not constant, under changing reading situations. Hence, different sets of stimuli, different response modalities, or different experimental setups yield new rules or different roles for already existing rules that were once thought to relate text to reader, as revealed in statistical interaction effects between interindividual factors, experimental factors or across experimental setups that were thought to tap into the very same reading process (Van Orden et al., 2001; Van Orden & Kloos, 2005). Instead, the language game hypothesis suggests that there is another property that is universal if one looks for the effect of text on readers, and that is to what degree a text has a structuring effect on a reader's behavior. This is not unlike the guiding intuition of current reading research, i.e., that text properties have a lawful relationship with reader behavior. However, it relaxes the requirements that objective text structures are always the source of order of a reader's behavior, and that objective text structures always act in the same, specific way on a reader's behavior. Understood this way, the language game hypothesis seems to hold two predictions that can be readily investigated in reading research: First, following Wittgenstein's analogy between rules in games and rules in language, one can propose that the degree of ruleabidingness with which a reader follows the relevant language rules within a specific reading situation is a measure of the competency of the reader with regard to the text at hand: The better a reader can follow the rules of a language game (i.e., a reading situation), the better he or she is able to ‘play that game’ (i.e., to read), that is to utilize relevant information in an efficient way during the course of reading. Hence, the degree of rule-abidingness of the reading process should be predictive of competency-related aspects, such as the level of reading skill, prior knowledge of the reader, text difficulty, and also situational factors such as attention and distraction (note, however, that the term ‘rule-abidingness’ is mainly used here in keeping with Wittgenstein's original terminology. In the next section, when linking language-use to dynamic systems, it will become clear that the term ‘rule-abidingness’ actually needs to be re-interpreted in terms of constraint, not rules per se).1 Second, and directly related to the question of specificity between text properties and cognition/reader behavior, one would expect that measures of the degree of rule-abidingness are more robust across different reading situations than specific measures that relate text properties to measures of reading behavior. For example, lexical variables (such as word frequency) show a strong influence on reading times in single-word reading tasks (such as naming or lexical decision), but bear little relation to reading times of connected text reading tasks (such as self-paced text reading; see Wallot et al., 2013; Wallot, O'Brien, Haussmann, Kloos, & Lyby, 2014). Also, the role of objective text properties is not the same across languages: For example, letter-sequence insensitivity (i.e., the ability of a reader to successfully read a word even though the
1 An anonymous reviewer of this manuscript pointed out that the equation of stronger rule-abidingness ¼ higher reading skill only pertains to comparisons within a particular reading task that can be interpreted as a language game (e.g., reading newspaper articles about economics). In order to assess general reading competency, it seems that one needs to assess different aspects of the reading process as well, for example how fast readers entrain to a new set of rules when switching from one language game to another e and across how many different language games they are able to do so.
Please cite this article in press as: Wallot, S., Understanding reading as a form of language-use: A language game hypothesis, New Ideas in Psychology (2015), http://dx.doi.org/10.1016/j.newideapsych.2015.07.006
S. Wallot / New Ideas in Psychology xxx (2015) 1e8
5
Fig. 1. Illustration of a time-series reflecting one of the dimensions of the Lorenz system (a) and a plot of the dynamics of all three equations of the Lorenz system in a 3-dimensional coordinate system (b). While the one-dimensional time-series in (a) gives the impression of a relatively irregular process, the 3-dimensional plot in (b) reveals that the behavior of the system exhibits a high degree of structure.
correct sequence of its letters has been reshuffled) is high in English, but low in Hebrew (Frost, 2012). Hence, if one wants to measure the competency of a reader across different reading tasks and languages, the text features that control the reading process will differ greatly across tasks and languages, but the relative degree of rule-abidingness might work the same way: As explained above, for the degree of rule-abidingness it is not important where the rules according to which a particular reading situation works come from, or what form they take. However, the better these rules are abided by, to use Wittgenstein's formulation again, the putatively better a reader functions in a particular reading situation. Hence, it is hypothesized that rule-abidingness will make for a more general measure of the reading process then the specific rules that can be observed to relate text to reader. Naturally, the question arises, how one can measure ruleabidingness in reading in such a general way. In the following section, I will present measures from dynamic systems analysis that lend themselves as good candidates to operationalize ruleabidingness in process measures of reading. Furthermore, I will argue that a usage-based understanding of reading, as exemplified by the concept of language games, can be very well interpreted in terms of basic concepts from dynamic systems, hence adding an appropriate couching of the proposed measures within the explanatory scheme of dynamic systems. 4. Language games as attractors Before we turn to the interpretation of language games as attractors of a dynamical system and derive appropriate measures,
we need to consider a first epistemological step regarding the quantification of rule-abidingness: Ideally, in order to quantify the degree of rule-abidingness of a process, we would need to know the rules that govern the process and possess a quantitative description of those rules. Again, lexical variables in single word reading are a good example here: It is known that the frequency of a word facilitates reading of that word, usually understood as a decrease in reading time of that word. However, as I have laid out above, one can often not trust that rules of one reading situation will transfer to another reading situation, or one might be interested in doing research on new reading situations, where the important variables are not known or not derivable from text properties. Hence, we need to find a logic of quantifying rule-abidingness of the reading process in the absence of knowledge about the rules. This can be done by adding the assumption that increases in rule-abidingness of a process lead to increases in structure of that process. That is, language is constraining behavioral options, not causing specific behaviors or mental contents (Ra˛ czaszek-Leonardi, 2009). This can be seen as a corollary of the statement that a process that evolved according to a set of rules is more lawful than a process that evolves randomly, in the absence of rules. Structure in turn implies that a process it not merely random. Given this assumption, one can measure the degree of structure that a process exhibits, and infer from it the degree of rule-abidingness of that process, or to be more precise, the degree to which that process is constrained. The problem is, of course, that structure is not always obvious. Structure can be hidden in nonlinearities, that make an actually ordered process seem random, or at least wildly fluctuating. Strange-attractors of dynamic systems are an example of
Please cite this article in press as: Wallot, S., Understanding reading as a form of language-use: A language game hypothesis, New Ideas in Psychology (2015), http://dx.doi.org/10.1016/j.newideapsych.2015.07.006
6
S. Wallot / New Ideas in Psychology xxx (2015) 1e8
such ‘hidden structures’. For example, consider the time-series in Fig. 1a. When one looks closely, one can see that the behavior of this time-series it not completely random, but for the most part, it looks like a fairly irregular process. The time-series is derived from the Lorenz system, a system of three coupled differential equations, whose behavior exhibits strange attractors. The time-series shown in Fig. 1a actually displays the behavior of one of the three equations of the Lorenz system. If we consider the behavior of all three of the equations together (Fig. 1b), then we can see that the seemingly irregular behavior of the time-series in Fig. 1a is actually part of rich, but ordered dynamics, exhibited by the Lorenz system (see Fig. 1b). The behaviors of the three variables are said to form an attractor, that is when one looks at Fig. 1b, one can see that the dynamics of the Lorenz system are confined (i.e., ‘attracted’) to a mere sub-space of values in the coordinate systems displayed in Fig. 1b. The coordinate values that are realized by the Lorenz system under a certain parametrization are more often encountered in some parts of the coordinate system, and are never encountered in other parts. Moreover, in the case of the Lorenz system we know that the dynamics are fully deterministic, that is, they fully ‘abide by rules’ e to stick with our language game terminology e that the coupled equations impose on them. Nevertheless, as an experimenter who has no access to the underlying causal relationships that govern, say, the reading process, one has to start with the irregularly looking time-series in Fig. 1a and ask the question, how rule-abiding e or to be more precise, how structured or constrained e the behavior of that timeseries is. This question can be answered empirically by employing time-series analysis techniques from the toolbox of complex systems analysis, such as Permutation entropy (Bandt & Pompe, 2002), Fractal analysis (Brown, & Leibovich, 2010), or Recurrence Quantification Analysis (Webber & Zbilut, 1994). All these techniques deliver measures that quantify the degree of order in a time-series. Let us, as an example, consider Recurrence Quantification Analysis (RQA): Simply put, RQA quantifies different aspects of how a time-
series shows repetitions. If these repetitions occur above chance, they will form greater patterns of repetitions, which are displayed on a Recurrence Plot (RP). Fig. 2a shows the RP of the time-series in Fig. 1a. The black structures mark periods of recurrent behavior in the time-series, while the white structures mark the absence of recurrences. What is notable is that recurrences are ordered in larger patterns, indicative of non-random behavior. There are several measures that can quantify the degree of this patterning of recurrences, one measure being %Determinism, which is the proportion of all diagonally adjacent recurrent structures divided by the sum of all observed recurrences in the RP. For Fig. 2a, % Determinism ¼ 99.54%, in contrast to Fig. 2b, which displays the RP of the shuffled time-series from Fig. 1a, where all temporal structure is removed by the shuffling process, and where %Determinism is consequently low, i.e., %Determinism ¼ 0.05%. This example illustrates how one can measure the structuredness of a signal without knowledge of the underlying processes that give rise to these structures. This analysis could be used on time-series of response time measures, for example to investigate the effects of text coherence on reading performance. To set up such an experiment, one could contrast four prominent reading conditions, using a) random word lists, b) random sentence lists, c) random paragraph lists, and d) connected text across several paragraphs. These four conditions will likely differ quantitatively in terms of the specific rules in terms of which the reading process can be described (i.e., decreasing effect strength of lexical variables on reading times; Wallot et al., 2013; 2014), and differ qualitatively in terms of the rules that can be observed (e.g., sentence wrap-up effects on reading times that can only be observed if the stimuli feature a clause or sentence structure, which is absent in random word lists; Haberlandt & Graesser, 1985). Finally, one might expect an increase in the structure of reading times, for example measured by RQA %Determinism, across the four tasks as additional layers of information that become available when moving from words to sentences, showing how these four language games differ in terms of the amount of constraint they put on the reading process e they are
Fig. 2. Recurrence plots (RPs) as examples of analyses of structure in time-series. Displayed are a) the RP and associated time-series for one dimension of the Lorenz system and b) the same time-series after the values have been shuffled. RPs are a 2-dimensional illustration of a 1-dimensional time-series. In an RP, dark areas mark periods of high recurrence, while white areas mark the absence of recurrence. When comparing RPs in a) and b), it becomes evident that the original time-series in a) shows a higher degree of recurrent structure, but also that the recurrent structure is ordered in larger patterns.
Please cite this article in press as: Wallot, S., Understanding reading as a form of language-use: A language game hypothesis, New Ideas in Psychology (2015), http://dx.doi.org/10.1016/j.newideapsych.2015.07.006
S. Wallot / New Ideas in Psychology xxx (2015) 1e8
increasingly constraining reading games. Beyond these practicalities, note that the conceptual context from which the measurements are derived (dynamical systems) mimic, or are at least a very good match with, the conceptual context from which the concept of language games stems (language-use). Both, dynamic systems theory and the language-use conception emphasize the temporal dimension of a process. Moreover, both emphasize the constitutive role of context for the details of a process: Just as the surface properties of a linguistic interaction between people are highly dependent on the shared common ground, the history between participants, and the topic at hand, dynamic systems emphasize the constitutive role of the parametrization of an equation system for the observable dynamics that result from this system: Different parametrizations of the equations can lead to qualitatively different behaviors. However, as long as the parameters are held stable, the observed behavior also yields some stable properties, which are captured by the term attractor. Hence, context in language-use shares with parametrizations in dynamic systems that both are slower time-scale properties that set the boundary conditions e constraints e of the linguistic/numerical behavior, and they are constitutive for the actual form that these behaviors take (Ra˛ czaszek-Leonardi, 2009). First studies on self-paced text reading and eye movements during reading using Recurrence and Fractal Analysis have shown that such dynamic measures make in fact for superior predictors of reading skill and reading comprehension compared to standard metrics such as reading speed (O'Brien, Wallot, Haussmann, & Kloos, 2014; Wallot et al., 2014; 2015). However, these studies have been conducted before the language game hypothesis was formulated, and thus do not provide a prospective test of the hypothesis' its predictions. 5. Conclusion In this article, I have tried to show that reading behavior can be understood as a form of language use (instead of a primarily perceptual-cognitive matching process), similar to other forms of natural language use, such as conversation. Specifically, I have proposed that the concepts of language games and form-of-life, originally formulated by Ludwig Wittgenstein (1953/2010), can provide a usage based perspective of language that is readily applicable to interpret reading behavior. Moreover, the concept of language games, when applied to the domain of reading, holds unique predictions for research in reading: Namely that the structuredness of process measures of reading is a reflection of reading competency, in the sense that reading is appropriately guided and constrained by the reading situation, and that structuredness measures will overall be less context dependent than specific measures of the reading process that map individual aspects of texts to individual aspects of behavior and cognition during reading. Furthermore, I have suggested that language games in reading can be interpreted as attractors in dynamical systems, where reading behavior is constrained and constituted by the slower-timescale aspects of a reading situation. Hence, it seems that dynamic measures, which are part of the statistical toolbox of dynamic systems analysis, will lend themselves to an operationalization of structuredness of the reading process. However, which of those measures will be most appropriate, or whether they will show converging results, is an open question e and similarly, whether such measures will in the end perform equally well or better to solve old and contemporary questions in reading research. In any case, the re-conceptualization of reading as language use opens up a broader perspective on reading that will permit linking research on reading conceptually, as well as methodologically, with other forms of natural language-use, such as conversation.
7
Author note I thank Anna Wallot for critical comments and helpful suggestions. References Bandt, C., & Pompe, B. (2002). Permutation entropy: a natural complexity measure for time series. Physical Review Letters, 88, 174102e1174102-2. Brown, C., & Liebovitch, L. (2010). Fractal analysis. Los Angeles, CA: Sage. Chafe, W., & Tannen, D. (1987). The relation between written and spoken language. Annual Review of Anthropology, 16, 383e407. Clark, H. H. (1996). Using language. Cambridge, MA: Cambridge University Press. Davidson, D. (1993). Locating literary language. In R. W. Dasenbrock (Ed.), Literary theory after Davidson (pp. 295e308). Penn State Press: University Park, PA. Frost, R. (2012). Towards a universal model of reading. Behavioural and Brain Science, 35, 263e279. Fusaroli, R., Bahrami, B., Olsen, K., Roepstorff, A., Rees, G., Frith, C., et al. (2012). Coming to terms: quantifying the benefits of linguistic coordination. Psychological Science, 23, 931e939. n, K. (2014). Dialog as interpersonal synFusaroli, R., Ra˛ czaszek-Leonardi, J., & Tyle ergy. New Ideas in Psychology, 32, 147e157. Graesser, A. C., & McNamara, D. S. (2011). Computational analyses of multilevel discourse comprehension. Topics in Cognitive Science, 3, 371e398. Grainger, J. (1990). Word frequency and neighborhood frequency effects in lexical decision and naming. Journal of Memory and Language, 29, 228e244. Grainger, J., & Jacobs, A. M. (1996). Orthographic processing in visual word recognition: a multiple read-out model. Psychological Review, 103, 518e565. Haberlandt, K. F., & Graesser, A. C. (1985). Component processes in text comprehension and some of their interactions. Journal of Experimental Psychology: General, 114, 357e374. Hohenstein, S., & Kliegl, R. (2014). Semantic preview benefit during reading. Journal of Experimental Psychology: Learning, Memory, and Cognition, 40, 166e190. Hunt, R. A., & Vipond, D. (1991). First, catch the rabbit: the methodological imperative and the dramatization of dialogic reading. Poetics, 20, 577e595. Keller, T. A., Carpenter, P. A., & Just, M. A. (2001). The neural bases of sentence comprehension: a fMRI examination of syntactic and lexical processing. Cerebral Cortex, 11, 223e237. Linell, P. (2005). The written language bias in Linguistics: Its nature, origins and transformations. London, UK: Routledge. McNerney, M. W., Goodwin, K. A., & Radvansky, G. A. (2011). A novel study: a situation model analysis of reading times. Discourse Processes, 48, 453e474. O'Brien, B. A., Wallot, S., Haussmann, A., & Kloos, H. (2014). Using complexity metrics to assess silent reading fluency: a cross-sectional study comparing oral and silent reading. Scientific Studies of Reading, 18, 235e254. Plamondon, R., & Srihari, S. N. (2000). Online and off-line handwriting recognition: a comprehensive survey. IEEE Transactions on Pattern Analysis and Machine Intelligence, 22, 63e84. Rayner, K., Chace, K. H., Slattery, T. J., & Ashby, J. (2006). Eye movements as reflections of comprehension processes in reading. Scientific Studies of Reading, 10, 241e255. Rayner, K., & Reichle, E. D. (2010). Models of the reading process. Cognitive Science, 1, 787e799. Reichle, E. D., Rayner, K., & Pollatsek, A. (2003). The EZ reader model of eyemovement control in reading: comparisons to other models. Behavioural and Brain Sciences, 26, 445e476. Reichle, E. D., Warren, T., & McConnell, K. (2009). Using E-Z reader to model the effects of higher-level language processing on eye movements during reading. Psychonomic Bulletin & Review, 16, 1e21. Richardson, D. C., Dale, R., & Kirkham, N. Z. (2007). The art of conversation is coordination common ground and the coupling of eye movements during dialogue. Psychological Science, 18, 407e413. Ra˛ czaszek-Leonardi, J. (2009). Symbols as constraints: the structuring role of dynamics and self-organization in natural language. Pragmatics & Cognition, 17, 653e676. Van Petten, C., & Kutas, M. (1991). Influences of semantic and syntactic context on open-and closed-class words. Memory & Cognition, 19, 95e112. Van Orden, G. C., & Kloos, H. (2005). The question of phonology and reading. In M. J. Snowling, & C. Hulme (Eds.), The science of reading: A handbook (pp. 61e78). Hoboken, NJ: Blackwell. Van Orden, G. C., Pennington, B. F., & Stone, G. O. (2001). What do double dissociations prove? Cognitive Science, 25, 111e172. Wallot, S. (2014). From “cracking the orthographic code” to “playing with language”: toward a usage-based foundation of the reading process. Frontiers in Psychology, 5. http://dx.doi.org/10.3389/fpsyg.2014.00891. Wallot, S., Hollis, G., & van Rooij, M. (2013). Connected text reading and differences in text reading fluency in adult readers. PLoS One, 8, e71914. Wallot, S., O'Brien, B. A., Coey, C. A., & Kelty-Stephen, D. (2015). Power-law fluctuations in eye movements predict text comprehension during connected text reading. In Proceedings of the 37th annual conference of the cognitive science society. Wallot, S., O'Brien, B. A., Haussmann, A., Kloos, A., & Lyby, M. S. (2014). The role of reading time complexity and reading speed in text comprehension. Journal of
Please cite this article in press as: Wallot, S., Understanding reading as a form of language-use: A language game hypothesis, New Ideas in Psychology (2015), http://dx.doi.org/10.1016/j.newideapsych.2015.07.006
8
S. Wallot / New Ideas in Psychology xxx (2015) 1e8
Experimental Psychology: Learning, Memory, and Cognition, 40, 1745e1765. Wallot, S., & Van Orden, G. (2011). Toward a lifespan metric of reading fluency. International Journal of Bifurcation and Chaos, 21, 1173e1192. Webber, C. L., & Zbilut, J. P. (1994). Dynamical assessment of physiological systems
and states using recurrence plot strategies. Journal of Applied Physiology, 76, 965e973. Wittgenstein, L. (1953/2010). Philosophical investigations. Hoboken, NJ: WileyBlackwell.
Please cite this article in press as: Wallot, S., Understanding reading as a form of language-use: A language game hypothesis, New Ideas in Psychology (2015), http://dx.doi.org/10.1016/j.newideapsych.2015.07.006
Contents lists available at ScienceDirect
New Ideas in Psychology journal homepage: www.elsevier.com/locate/newideapsych
Understanding reading as a form of language-use: A language game hypothesis Sebastian Wallot* Interacting Minds Centre, Department of Culture and Society, Aarhus University, Denmark
a r t i c l e i n f o
a b s t r a c t
Article history: Received 15 March 2015 Received in revised form 17 July 2015 Accepted 20 July 2015 Available online xxx
Reading research and research on conversation have followed different paths: While the research program for reading committed itself to a relatively static view of language, where objective text properties serve to elicit specific effects on cognition and behavior of a reader, research on conversation has embraced a language-use perspective, where language is primarily seen as a dynamic, context dependent process. In this essay I contrast these two perspectives, and argue that in order to reach a unified understanding of natural language e be it reading, talking, or conversing e one needs to adopt a languageuse perspective. Furthermore, I describe how reading can be seen as a form of language-use, and how the current landscape of research on reading can be re-interpreted in terms of a dynamic, context-sensitive perspective on language. In particular, I propose that the concept of ‘language games’ serves as a good starting point to conceive reading as a form of language-use, describe how one can derive first concrete hypotheses by re-interpreting reading in terms of language games, and show how they can be readily operationalized using tools from dynamic systems analysis. © 2015 Elsevier Ltd. All rights reserved.
Keywords: Reading Conversation Language-use Dynamic systems Language games
1. Introduction Reading is a cultural-cognitive skill that most people exercise on a daily basis. From a psychological perspective, reading is interesting because of this cognitive-cultural aspect: As a cultural skill, reading is of practical relevance for education, economy, and recreation. As a cognitive skill, reading touches many aspects of human cognition and interaction. Reading is a form of natural languageuse, just as talking, listening, conversing. It is communication, understanding, thinking, and perception. Reading has evolved for communicative purposes, to pass along messages, but also to extend the mind into the environment, for example to enhance memory by writing notes or to reflect upon one's own thoughts by writing them down. Obviously, reading mandates perception, as text is a property of the environment, and needs to be seen in order to be utilized. However, the perceptual side of reading has received much more scientific attention than any of its other aspects. The main puzzle that psychologists sought to solve for reading is, how visual contrast gradients in the environment can be identified as words/
* Interacting Minds Centre, Aarhus University Jens Chr. Skous Vej 4, Building 1.483, 3rd floor, 8000 Aarhus C, Denmark. E-mail address: [email protected].
language by readers. Language can be written/printed in many diverse and unique ways, and success with extracting language from an environmental sources has been limited to clean print, or isolated letters (Plamondon & Srihari, 2000). Hence, research has taken a half-step back from the straight question of how visual features are recognized by a reader, and settled instead on linguistic features that help a reader to recognize a particular word of written language. Such features, called lexical variables, are the centerpiece of the well-developed contemporary theories and models of reading, implemented in computational models of word recognition and models of eye-movement control during reading (e.g., Grainger & Jacobs, 1996; Reichle, Rayner, & Pollatsek, 2003). Lexical variables are quantifiable aspects of a word, usually based on corpus-linguistic analyses. One of the most prominent lexical variables is called word frequency. In order to quantify the frequency of a word, one counts how often that word appears in a representative corpus of texts. The relative number of appearances of that word in the corpus is then an estimator of its frequency. Quantitative measures of word frequency can be correlated with quantitative measures of the reading process (e.g., reaction times, fixation durations …), and further refinements can be made, for example the prominent logarithmic transformation of the frequency values, or theoretical parameterization of a regression model that converts frequencies into gains in terms of reading
http://dx.doi.org/10.1016/j.newideapsych.2015.07.006 0732-118X/© 2015 Elsevier Ltd. All rights reserved.
Please cite this article in press as: Wallot, S., Understanding reading as a form of language-use: A language game hypothesis, New Ideas in Psychology (2015), http://dx.doi.org/10.1016/j.newideapsych.2015.07.006
2
S. Wallot / New Ideas in Psychology xxx (2015) 1e8
speed in milliseconds. The basic idea is that words exhibit stable linguistic characteristics that are captured by lexical variables, which in turn allow a reader to reliably identify words, the assumed building blocks of written language. A second strand of research that is also relatively prominent in reading research is concerned with comprehension processes of the reader (e.g., Graesser & McNamara, 2011). Interestingly, the research that focuses on the perceptual front-end of reading and the research that focuses on later-stage comprehension processes is relatively disjoint, with the former focusing heavily on process measures of reading, and the latter focusing on outcome measures of reading (but there are ongoing attempts to bring results from comprehension research into process measures of reading e Reichle, Warren, & McConnell, 2009). Conceptually, however, comprehension research is well aligned at its meta-theoretical core with the perceptual/process research on reading. Just as the perceptual/process research on reading assumes that a particular word exhibits particular features that lead to particular effects on the side of the reader, comprehension focused research assumes that comprehension of a text implies comprehension of something particular. That is, that there are stable features, for example on the sentence level, that act either as a comprehension-marker of that sentence or that connect two adjacent sentences (e.g., situation model dimensions), and that those features are reliably used by readers to comprehend a particular message (cf. Graesser & McNamara, 2011). To summarize, reading research e on the level of perception and on the level of comprehension e has thrived on the assumption that the reading process is specific. It is specific in the sense that there are specific aspects of texts (e.g., certain lexical characteristics of words) that map onto specific cognitive processes (e.g., word identification processes) and specific cognitive contents (e.g., the meanings of words as prescribed in their representation in the mental lexicon). In order to be specific in this way, there must be a stable relationship between aspects of the text and responses of a reader with regard to the text. One can only establish a link between, say, a word property and a particular mental content that in turn triggers a measureable response (e.g., a reaction time, a fixation duration, an event-related potential) if both sides on this mapping relationship remain stable. Hence, lexical variables such as word frequency are only useful as universtal building blocks for theories of reading if the frequency of a word is relatively constant, and if the meaning of a word remains more or less the same across different reading situations. If, for example, the meaning of a word can change radically or instantaneously in different reading situations, then word frequency would not be of any help to link the visual impression of a word to its meaning. The frequency of a word would not be specific any more with regard to the particular mental representation it supposedly helps to index. The guiding assumption of specific relations between text properties and cognitive processes/contents has let to a fruitful e maybe even the most fruitful e research program in experimental psychology, and has generated practical results that have spilled over in the world of every-day readers, such as the automated readability metrics that are now part of many word processing programs (e.g., the Flesh-Kincaid index), or the design of speedreading software (e.g., implementation of the optimal viewing position effect in RSVP reading software). At the same time, however, research on reading has lived a somewhat exclusive life in its own traditional niche, apart from other natural language activities, especially conversation. Some researchers have tried to establish a conceptual link between conversation/spoken language and reading (Chafe & Tannen, 1987), but empirical research on the topic is virtually absent. The main reason why it seems so difficult to reconcile reading with
conversation seems to be the diverging conception in these two research traditions with regard to what language is, and how it inherently works. Compared to research on reading, which has focused on perception and the mapping of stable word features to cognition, research on conversation, communication and social interaction stands on the other end of the spectrum. Research on reading might have taken the view of reading as a strongly componential process as a result of mainly dealing with pieces of written language, which seem to convey a static picture of language because that is how texts make language appear, exhibiting its own version of the what has been called the written language bias (Linell, 2005) in linguistics. In contrast, research on conversation has adopted a perspective with different emphasis, namely how its bearers use language. Note that this ontological divergence in the conception of language should strike one as odd, as a naïve perspective on language development would suggest that the two are closely related: After all, reading always develops on the language capabilities that are based on conversation and social interaction in infancy and childhood. So intuitively, these two domains of language should not work according to separate principles, but share a common core. Compared to the componential/perceptual stance adopted by reading research, research on conversation has adopted a usagebased stance, which offers a strongly contextualized perspective on language e i.e., that it is not a set of (perceptual mapping) rules that structure language, but that language is structured by functional aspects that differ from context to context in which it is used. Among other things, this reserves a fundamental role for the intentions of participants in a communication, for the social context and purpose of the communicative situation, as well as for the dynamics of conversation e that is, that communication is not built out of static components, but that the development of a communicative process is consequential for the outcome of that commun, 2014). nication (Fusaroli, Ra˛ czaszek-Leonardi, & Tyle Admittedly, all these considerations seem to be fine, if not inevitable, for such a thing as two people standing in the same room, being able to talk to each other, see and hear each other, and interact with each other. However, from the surface, reading seems quite different, where a single person sits or stands still, scanning the environment for patterns. So how does all of this pertain to reading at all? And why should one even bother? 2. Reading is complex, not just complicated Perhaps the reason why one regards the phenomenon 'conversation' as being appropriately described by an account that allows central roles for intentions, context and dynamics and shies away from an all too strict reductionism is, because conversation is easily seen as a complex phenomenon: Research on language in social interaction (Clark, 1996), as well as common everyday experience suggest that an abundance of interdependent factors play a role for conversation, and that conversation can either go wrong or succeed in unexpected ways, no matter how clear the goals of the interlocutors or their communicative ability, or how seemingly simple the topic. For example in a perceptual identification task, two participants see each the same six striped patterns on a computer monitor, one being slightly different from the other five. They are to decide which of the six is the odd one and individually submit their decision to the experimenter. If the two do not agree, then they are allowed to take a break and discuss which choice they should go with. Interestingly, each pair develops its very own vocabulary to deal with this kind of task (Fusaroli et al., 2012). For example, two participants developed a cheese-confidence-scale, where each participant told the other how confident they were that their choice was correct in ascending orders of cheese-stinkyness (“Cheese. It's
Please cite this article in press as: Wallot, S., Understanding reading as a form of language-use: A language game hypothesis, New Ideas in Psychology (2015), http://dx.doi.org/10.1016/j.newideapsych.2015.07.006
S. Wallot / New Ideas in Psychology xxx (2015) 1e8
cheese. It's stinky cheese. Stilton!” e Riccardo Fusaroli, personal communication). This odd and humorous example should simply serve to illustrate that successful communication during conversation can take an abundance of forms, and it is hard to pin them down to a specific corpus of words that would, for example, delimit good from bad interactions. It is rather the functionality that is driving conversational language choice, not a stereotyped lexicon. However, when one thinks about reading, it becomes clear that there are similarities between reading and conversation beyond the emphasis of the perceptual level: Just as communication in a successful conversation depends on mutual consideration of the intentions of interlocutors, reading depends on the intentions of the reader. Readers' intentions have consequences for how reading unfolds over time, for reading behavior, emotional experience, and memory (Hunt & Vipond, 1991). Furthermore, reading also depends on the (assumed and real) intentions of the author, who becomes the equivalent of an interlocutor in a real conversation (Davidson, 1993). Again, however, the role of the author is not static: As the text unfolds, the author's intentions need to be revised, such as when irony is used, or the reader is ‘led down the wrong path’. Just as in real conversations, where interlocutors go back and forth to establish common ground or pause for reflection (Fusaroli et al., 2012), research on reading has observed feedback loops during reading, such as re-reading previous text passages (Rayner, Chace, Slattery, & Ashby, 2006), a sort of substitute for online conversational exchange. Hence, seen this way, there is a substantial overlap between how conversation and reading work, and as one approaches naturalistic texts e as opposed to, say, random word lists e these factors, which are hard to pinpoint to objective text features, begin to matter more and more. In addition to this more phenomenological side, reading has also proved to be complex with regard to the strong commitment towards its anchoring on linguistic properties, such as lexical variables, syntax, and text-structure based measures of coherence and semantics. If one considers the simplest reading tasks, such as lexical decision or word naming, where lexical word features do play a substantial role, the complexity of reading reveals itself through nonlinear correlations between the different lexical features that are thought to drive reading, as well as through the effects of these features on reading performance: In general, the majority of lexical features (such as word frequency, word length, co-occurrence, neighborhood frequency etc.) are highly correlated amongst each other. Moreover, the correlations among these features are not linear with regard to reading performance: For example, the effects of word frequency depend on the semantic context within which a word is presented (Van Petten & Kutas, 1991), or orthographic neighborhood frequency (Grainger, 1990). The heterogeneity of findings in reading is a problem because this makes it difficult to say which role a particular factor that is investigated in reading research actually plays for the reading process e and this problem is not limited to the level of word reading. Prominent sentence-reading tasks reveal that the role of word properties is dependent on syntactic complexity (Keller, Carpenter, & Just, 2001). Moreover, eye-movement studies during reading have generally called into question in how far individual words can be treated as separable entities: When a reader can see more than one word at a time, words at the periphery of vision are already being pre-processed by the reader (Hohenstein & Kliegl, 2014), and this parafoveal preview influences the reading of the word that is currently being fixated by the reader, generating facilitative or inhibitory effects for the reading process depending on the relation between these two words. As we scale up once more from sentence to connected text reading, lexical features loose their constitutive role for the reading process all together (Wallot, Hollis, & van Rooij, 2013). Similarly,
3
text coherence metrics that were developed to characterize text reading fail to yield many of their predicted effects on a reader's performance, if one observes reading of long connected text that spans more than a few paragraphs (McNerney, Goodwin, & Radvansky, 2011). One can say that it is in fact not even clear what drives the reading process of naturalistics text, as it seems that many of the relatively simple relations between objective text features and reader performance that are observed in simpler reading tasks fail to reveal themselves in naturalistic reading. It is almost as if naturalistic reading would be an entirely different phenomenon, at least something else than a concatenation of words or sentences. Finally, the different kinds of reading tasks that have been used to measure reading in psychological laboratories add another layer of complexity on top of what different kinds of text materials provide: Reading word lists silently versus aloud already changes the way that reading performance and text properties relate (Grainger, 1990), and such interactions between task performance and context seem to be the rule, rather than the exception (Van Orden, Pennington, & Stone, 2001). The above examples are meant to illustrate that reading is not merely complicated, but complex. If reading were only complicated, then reading research is on the right track and by describing a great number of relevant factors, a relatively simple and coherent theory of reading would stand at the end of those efforts. However, such a theory is nowhere in sight (Rayner & Reichle, 2010) and it seems that reading is more than just complicated. The factors that are investigated in reading are intricately intertwined with each other, and change their role and relation as linguistic, but also environmental aspects of a reading situation are changed. Hence, reading is not just complicated, it is complex, and with the examples above I wanted to illustrate how some of the assumptions of a language-use perspective readily also pertain to reading research that has largely been conducted under different premises. In particular, it seems that the presence of an author that provides for an intentional background of a reading task changes how reading unfolds, something which is (largely) absent in the prominent isolated word and sentence reading tasks e a random word list is very hard to read as an intended message. Moreover, non-linguistic aspects seem to highly influence the reading process as well, such as how a text is displayed (one word at a time, one sentence at a time, one page at a time e Wallot & Van Orden, 2011), what tasks and measurement setups are utilized (Grainger, 1990), and what situation a participant is asked to perform the reading in (Hunt & Vipond, 1991). Here, one might see traces of what have been thought to be genuine conversational properties e intention, situation, and non-linguistic context e in reading. That is, effects of things that are themselves not textual entities, but co-determine how language is perceived and understood by a reader. The question is, what would be an adequate framework for conceiving the situation of a text and a reader in terms of languageuse? Such a framework should be connectable to conversation, but also provide initial hypotheses and an adequate measurementframework for empirical investigations of the reading process. After all, prominent measures that capture aspects of language-use in studies on conversation, such as coupling of movements (Richardson, Dale, & Kirkham, 2007) or terminology (Fusaroli et al., 2012) between participants, cannot be obtained from a reading situation. In the next section, I will outline such a framework, based on the concept of language games, originally stipulated by Ludwig Wittgenstein in the ‘Philosophical Investigations’, and show how analytic tools from the toolbox of dynamic systems can be used to operationalize these concepts.
Please cite this article in press as: Wallot, S., Understanding reading as a form of language-use: A language game hypothesis, New Ideas in Psychology (2015), http://dx.doi.org/10.1016/j.newideapsych.2015.07.006
4
S. Wallot / New Ideas in Psychology xxx (2015) 1e8
3. Conceiving reading as language use As I have proposed elsewhere (Wallot, 2014), the concepts of form-of-life and language games (Wittgenstein, 1953/2010) might serve as a theoretical anchor for a unifying, usage-based perspective on conversation and reading, with implications for how research on reading as language-use can be done. Both of these concepts were most likely formulated with conversational situations in mind, not with reading. However, as I will explain in the following, these two concepts lend themselves well to an application of the reading situation. In the ‘Philosophical Investigations’, linguistic structure is described as a variant phenomenon, and contrary to the dominant linguistic perspective, objective structures in language (such as syntax) are not primarily seen as building blocks of language, its production, or comprehension, but as emergent phenomena. This follows from the two related concepts, form-of-life and language games: Form-of-life describes the idea that language is always used within a set of (non-linguistic) constraints that shape and constitute the contexts within which people use language. These constraints can span historical influences, geographical and material aspects of the current environment in which a conversation takes place, as well as the social norms and personal characteristics and relationships of the conversants. Form-of-life supplies the repertoire of potentially meaningful language-uses. Within form-of-life, different language games can be played. Language games can be seen as a concrete realization from the repertoire supplied by form-of-life. They are an actual communicative exchange that can be observed. Wittgenstein (1953/2010) chose the term game, because a concrete instance of a conversation looks like a rule-abiding activity that is functional towards keeping the communicative exchange going. However, just as there are many different games with many different rules, the goals and nature of communicative situations are very diverse. Moreover, as the nature of a communicative situation changes, so do the rules according to which that communication unfolds. Hence, the rules that are observed in the language between two interlocutors are not a reflection of the principal architecture of language, but they are themselves emergent, being realized within the constraints of form-of-life, the actual situation in which that communication takes place. Let us consider the application of form-of-life and language games to reading, and preliminary hypotheses for reading research that can be derived from these concepts: First of all, the language game perspective shares a sentiment regarding the effects of language on a person's (linguistic) behavior with classical reading research, namely that language-use e reading e is a rule-abiding activity. However, for the majority of reading research, this has translated into a stricter assumption, namely that reading can be defined as a specific reaction of the reader to specific aspects of objective text structure. That is, the structure observed in reading behavior is to be traced back to objective text structures (lexical variables, syntactical features …), and an understanding of these structures would mean to understand the universal rules by which reading behavior abides. Moreover, the relation between text and reader can be described by specific rules (i.e., the higher the frequency of a word, the shorter the fixation duration of that word), which would allow to predict a reader's behavior across any reading situation. In the conception of language games, rules play a different role: Reading is also seen as a structured activity, and in a particular reading situation (e.g., reading random word lists and responding to each word with a button press), one can observe and formulate particular rules according to which the observed reading process behaves. However, these rules are not universal, and are themselves
a result of the reading situation. As one changes the reading situation, that is, changes the language game, partially or completely different rules might be observed that guide a reader's behavior. This seems also consonant with the observation that any particular rule that has been formulated by reading researchers to govern the reading process has been observed to be variant, not constant, under changing reading situations. Hence, different sets of stimuli, different response modalities, or different experimental setups yield new rules or different roles for already existing rules that were once thought to relate text to reader, as revealed in statistical interaction effects between interindividual factors, experimental factors or across experimental setups that were thought to tap into the very same reading process (Van Orden et al., 2001; Van Orden & Kloos, 2005). Instead, the language game hypothesis suggests that there is another property that is universal if one looks for the effect of text on readers, and that is to what degree a text has a structuring effect on a reader's behavior. This is not unlike the guiding intuition of current reading research, i.e., that text properties have a lawful relationship with reader behavior. However, it relaxes the requirements that objective text structures are always the source of order of a reader's behavior, and that objective text structures always act in the same, specific way on a reader's behavior. Understood this way, the language game hypothesis seems to hold two predictions that can be readily investigated in reading research: First, following Wittgenstein's analogy between rules in games and rules in language, one can propose that the degree of ruleabidingness with which a reader follows the relevant language rules within a specific reading situation is a measure of the competency of the reader with regard to the text at hand: The better a reader can follow the rules of a language game (i.e., a reading situation), the better he or she is able to ‘play that game’ (i.e., to read), that is to utilize relevant information in an efficient way during the course of reading. Hence, the degree of rule-abidingness of the reading process should be predictive of competency-related aspects, such as the level of reading skill, prior knowledge of the reader, text difficulty, and also situational factors such as attention and distraction (note, however, that the term ‘rule-abidingness’ is mainly used here in keeping with Wittgenstein's original terminology. In the next section, when linking language-use to dynamic systems, it will become clear that the term ‘rule-abidingness’ actually needs to be re-interpreted in terms of constraint, not rules per se).1 Second, and directly related to the question of specificity between text properties and cognition/reader behavior, one would expect that measures of the degree of rule-abidingness are more robust across different reading situations than specific measures that relate text properties to measures of reading behavior. For example, lexical variables (such as word frequency) show a strong influence on reading times in single-word reading tasks (such as naming or lexical decision), but bear little relation to reading times of connected text reading tasks (such as self-paced text reading; see Wallot et al., 2013; Wallot, O'Brien, Haussmann, Kloos, & Lyby, 2014). Also, the role of objective text properties is not the same across languages: For example, letter-sequence insensitivity (i.e., the ability of a reader to successfully read a word even though the
1 An anonymous reviewer of this manuscript pointed out that the equation of stronger rule-abidingness ¼ higher reading skill only pertains to comparisons within a particular reading task that can be interpreted as a language game (e.g., reading newspaper articles about economics). In order to assess general reading competency, it seems that one needs to assess different aspects of the reading process as well, for example how fast readers entrain to a new set of rules when switching from one language game to another e and across how many different language games they are able to do so.
Please cite this article in press as: Wallot, S., Understanding reading as a form of language-use: A language game hypothesis, New Ideas in Psychology (2015), http://dx.doi.org/10.1016/j.newideapsych.2015.07.006
S. Wallot / New Ideas in Psychology xxx (2015) 1e8
5
Fig. 1. Illustration of a time-series reflecting one of the dimensions of the Lorenz system (a) and a plot of the dynamics of all three equations of the Lorenz system in a 3-dimensional coordinate system (b). While the one-dimensional time-series in (a) gives the impression of a relatively irregular process, the 3-dimensional plot in (b) reveals that the behavior of the system exhibits a high degree of structure.
correct sequence of its letters has been reshuffled) is high in English, but low in Hebrew (Frost, 2012). Hence, if one wants to measure the competency of a reader across different reading tasks and languages, the text features that control the reading process will differ greatly across tasks and languages, but the relative degree of rule-abidingness might work the same way: As explained above, for the degree of rule-abidingness it is not important where the rules according to which a particular reading situation works come from, or what form they take. However, the better these rules are abided by, to use Wittgenstein's formulation again, the putatively better a reader functions in a particular reading situation. Hence, it is hypothesized that rule-abidingness will make for a more general measure of the reading process then the specific rules that can be observed to relate text to reader. Naturally, the question arises, how one can measure ruleabidingness in reading in such a general way. In the following section, I will present measures from dynamic systems analysis that lend themselves as good candidates to operationalize ruleabidingness in process measures of reading. Furthermore, I will argue that a usage-based understanding of reading, as exemplified by the concept of language games, can be very well interpreted in terms of basic concepts from dynamic systems, hence adding an appropriate couching of the proposed measures within the explanatory scheme of dynamic systems. 4. Language games as attractors Before we turn to the interpretation of language games as attractors of a dynamical system and derive appropriate measures,
we need to consider a first epistemological step regarding the quantification of rule-abidingness: Ideally, in order to quantify the degree of rule-abidingness of a process, we would need to know the rules that govern the process and possess a quantitative description of those rules. Again, lexical variables in single word reading are a good example here: It is known that the frequency of a word facilitates reading of that word, usually understood as a decrease in reading time of that word. However, as I have laid out above, one can often not trust that rules of one reading situation will transfer to another reading situation, or one might be interested in doing research on new reading situations, where the important variables are not known or not derivable from text properties. Hence, we need to find a logic of quantifying rule-abidingness of the reading process in the absence of knowledge about the rules. This can be done by adding the assumption that increases in rule-abidingness of a process lead to increases in structure of that process. That is, language is constraining behavioral options, not causing specific behaviors or mental contents (Ra˛ czaszek-Leonardi, 2009). This can be seen as a corollary of the statement that a process that evolved according to a set of rules is more lawful than a process that evolves randomly, in the absence of rules. Structure in turn implies that a process it not merely random. Given this assumption, one can measure the degree of structure that a process exhibits, and infer from it the degree of rule-abidingness of that process, or to be more precise, the degree to which that process is constrained. The problem is, of course, that structure is not always obvious. Structure can be hidden in nonlinearities, that make an actually ordered process seem random, or at least wildly fluctuating. Strange-attractors of dynamic systems are an example of
Please cite this article in press as: Wallot, S., Understanding reading as a form of language-use: A language game hypothesis, New Ideas in Psychology (2015), http://dx.doi.org/10.1016/j.newideapsych.2015.07.006
6
S. Wallot / New Ideas in Psychology xxx (2015) 1e8
such ‘hidden structures’. For example, consider the time-series in Fig. 1a. When one looks closely, one can see that the behavior of this time-series it not completely random, but for the most part, it looks like a fairly irregular process. The time-series is derived from the Lorenz system, a system of three coupled differential equations, whose behavior exhibits strange attractors. The time-series shown in Fig. 1a actually displays the behavior of one of the three equations of the Lorenz system. If we consider the behavior of all three of the equations together (Fig. 1b), then we can see that the seemingly irregular behavior of the time-series in Fig. 1a is actually part of rich, but ordered dynamics, exhibited by the Lorenz system (see Fig. 1b). The behaviors of the three variables are said to form an attractor, that is when one looks at Fig. 1b, one can see that the dynamics of the Lorenz system are confined (i.e., ‘attracted’) to a mere sub-space of values in the coordinate systems displayed in Fig. 1b. The coordinate values that are realized by the Lorenz system under a certain parametrization are more often encountered in some parts of the coordinate system, and are never encountered in other parts. Moreover, in the case of the Lorenz system we know that the dynamics are fully deterministic, that is, they fully ‘abide by rules’ e to stick with our language game terminology e that the coupled equations impose on them. Nevertheless, as an experimenter who has no access to the underlying causal relationships that govern, say, the reading process, one has to start with the irregularly looking time-series in Fig. 1a and ask the question, how rule-abiding e or to be more precise, how structured or constrained e the behavior of that timeseries is. This question can be answered empirically by employing time-series analysis techniques from the toolbox of complex systems analysis, such as Permutation entropy (Bandt & Pompe, 2002), Fractal analysis (Brown, & Leibovich, 2010), or Recurrence Quantification Analysis (Webber & Zbilut, 1994). All these techniques deliver measures that quantify the degree of order in a time-series. Let us, as an example, consider Recurrence Quantification Analysis (RQA): Simply put, RQA quantifies different aspects of how a time-
series shows repetitions. If these repetitions occur above chance, they will form greater patterns of repetitions, which are displayed on a Recurrence Plot (RP). Fig. 2a shows the RP of the time-series in Fig. 1a. The black structures mark periods of recurrent behavior in the time-series, while the white structures mark the absence of recurrences. What is notable is that recurrences are ordered in larger patterns, indicative of non-random behavior. There are several measures that can quantify the degree of this patterning of recurrences, one measure being %Determinism, which is the proportion of all diagonally adjacent recurrent structures divided by the sum of all observed recurrences in the RP. For Fig. 2a, % Determinism ¼ 99.54%, in contrast to Fig. 2b, which displays the RP of the shuffled time-series from Fig. 1a, where all temporal structure is removed by the shuffling process, and where %Determinism is consequently low, i.e., %Determinism ¼ 0.05%. This example illustrates how one can measure the structuredness of a signal without knowledge of the underlying processes that give rise to these structures. This analysis could be used on time-series of response time measures, for example to investigate the effects of text coherence on reading performance. To set up such an experiment, one could contrast four prominent reading conditions, using a) random word lists, b) random sentence lists, c) random paragraph lists, and d) connected text across several paragraphs. These four conditions will likely differ quantitatively in terms of the specific rules in terms of which the reading process can be described (i.e., decreasing effect strength of lexical variables on reading times; Wallot et al., 2013; 2014), and differ qualitatively in terms of the rules that can be observed (e.g., sentence wrap-up effects on reading times that can only be observed if the stimuli feature a clause or sentence structure, which is absent in random word lists; Haberlandt & Graesser, 1985). Finally, one might expect an increase in the structure of reading times, for example measured by RQA %Determinism, across the four tasks as additional layers of information that become available when moving from words to sentences, showing how these four language games differ in terms of the amount of constraint they put on the reading process e they are
Fig. 2. Recurrence plots (RPs) as examples of analyses of structure in time-series. Displayed are a) the RP and associated time-series for one dimension of the Lorenz system and b) the same time-series after the values have been shuffled. RPs are a 2-dimensional illustration of a 1-dimensional time-series. In an RP, dark areas mark periods of high recurrence, while white areas mark the absence of recurrence. When comparing RPs in a) and b), it becomes evident that the original time-series in a) shows a higher degree of recurrent structure, but also that the recurrent structure is ordered in larger patterns.
Please cite this article in press as: Wallot, S., Understanding reading as a form of language-use: A language game hypothesis, New Ideas in Psychology (2015), http://dx.doi.org/10.1016/j.newideapsych.2015.07.006
S. Wallot / New Ideas in Psychology xxx (2015) 1e8
increasingly constraining reading games. Beyond these practicalities, note that the conceptual context from which the measurements are derived (dynamical systems) mimic, or are at least a very good match with, the conceptual context from which the concept of language games stems (language-use). Both, dynamic systems theory and the language-use conception emphasize the temporal dimension of a process. Moreover, both emphasize the constitutive role of context for the details of a process: Just as the surface properties of a linguistic interaction between people are highly dependent on the shared common ground, the history between participants, and the topic at hand, dynamic systems emphasize the constitutive role of the parametrization of an equation system for the observable dynamics that result from this system: Different parametrizations of the equations can lead to qualitatively different behaviors. However, as long as the parameters are held stable, the observed behavior also yields some stable properties, which are captured by the term attractor. Hence, context in language-use shares with parametrizations in dynamic systems that both are slower time-scale properties that set the boundary conditions e constraints e of the linguistic/numerical behavior, and they are constitutive for the actual form that these behaviors take (Ra˛ czaszek-Leonardi, 2009). First studies on self-paced text reading and eye movements during reading using Recurrence and Fractal Analysis have shown that such dynamic measures make in fact for superior predictors of reading skill and reading comprehension compared to standard metrics such as reading speed (O'Brien, Wallot, Haussmann, & Kloos, 2014; Wallot et al., 2014; 2015). However, these studies have been conducted before the language game hypothesis was formulated, and thus do not provide a prospective test of the hypothesis' its predictions. 5. Conclusion In this article, I have tried to show that reading behavior can be understood as a form of language use (instead of a primarily perceptual-cognitive matching process), similar to other forms of natural language use, such as conversation. Specifically, I have proposed that the concepts of language games and form-of-life, originally formulated by Ludwig Wittgenstein (1953/2010), can provide a usage based perspective of language that is readily applicable to interpret reading behavior. Moreover, the concept of language games, when applied to the domain of reading, holds unique predictions for research in reading: Namely that the structuredness of process measures of reading is a reflection of reading competency, in the sense that reading is appropriately guided and constrained by the reading situation, and that structuredness measures will overall be less context dependent than specific measures of the reading process that map individual aspects of texts to individual aspects of behavior and cognition during reading. Furthermore, I have suggested that language games in reading can be interpreted as attractors in dynamical systems, where reading behavior is constrained and constituted by the slower-timescale aspects of a reading situation. Hence, it seems that dynamic measures, which are part of the statistical toolbox of dynamic systems analysis, will lend themselves to an operationalization of structuredness of the reading process. However, which of those measures will be most appropriate, or whether they will show converging results, is an open question e and similarly, whether such measures will in the end perform equally well or better to solve old and contemporary questions in reading research. In any case, the re-conceptualization of reading as language use opens up a broader perspective on reading that will permit linking research on reading conceptually, as well as methodologically, with other forms of natural language-use, such as conversation.
7
Author note I thank Anna Wallot for critical comments and helpful suggestions. References Bandt, C., & Pompe, B. (2002). Permutation entropy: a natural complexity measure for time series. Physical Review Letters, 88, 174102e1174102-2. Brown, C., & Liebovitch, L. (2010). Fractal analysis. Los Angeles, CA: Sage. Chafe, W., & Tannen, D. (1987). The relation between written and spoken language. Annual Review of Anthropology, 16, 383e407. Clark, H. H. (1996). Using language. Cambridge, MA: Cambridge University Press. Davidson, D. (1993). Locating literary language. In R. W. Dasenbrock (Ed.), Literary theory after Davidson (pp. 295e308). Penn State Press: University Park, PA. Frost, R. (2012). Towards a universal model of reading. Behavioural and Brain Science, 35, 263e279. Fusaroli, R., Bahrami, B., Olsen, K., Roepstorff, A., Rees, G., Frith, C., et al. (2012). Coming to terms: quantifying the benefits of linguistic coordination. Psychological Science, 23, 931e939. n, K. (2014). Dialog as interpersonal synFusaroli, R., Ra˛ czaszek-Leonardi, J., & Tyle ergy. New Ideas in Psychology, 32, 147e157. Graesser, A. C., & McNamara, D. S. (2011). Computational analyses of multilevel discourse comprehension. Topics in Cognitive Science, 3, 371e398. Grainger, J. (1990). Word frequency and neighborhood frequency effects in lexical decision and naming. Journal of Memory and Language, 29, 228e244. Grainger, J., & Jacobs, A. M. (1996). Orthographic processing in visual word recognition: a multiple read-out model. Psychological Review, 103, 518e565. Haberlandt, K. F., & Graesser, A. C. (1985). Component processes in text comprehension and some of their interactions. Journal of Experimental Psychology: General, 114, 357e374. Hohenstein, S., & Kliegl, R. (2014). Semantic preview benefit during reading. Journal of Experimental Psychology: Learning, Memory, and Cognition, 40, 166e190. Hunt, R. A., & Vipond, D. (1991). First, catch the rabbit: the methodological imperative and the dramatization of dialogic reading. Poetics, 20, 577e595. Keller, T. A., Carpenter, P. A., & Just, M. A. (2001). The neural bases of sentence comprehension: a fMRI examination of syntactic and lexical processing. Cerebral Cortex, 11, 223e237. Linell, P. (2005). The written language bias in Linguistics: Its nature, origins and transformations. London, UK: Routledge. McNerney, M. W., Goodwin, K. A., & Radvansky, G. A. (2011). A novel study: a situation model analysis of reading times. Discourse Processes, 48, 453e474. O'Brien, B. A., Wallot, S., Haussmann, A., & Kloos, H. (2014). Using complexity metrics to assess silent reading fluency: a cross-sectional study comparing oral and silent reading. Scientific Studies of Reading, 18, 235e254. Plamondon, R., & Srihari, S. N. (2000). Online and off-line handwriting recognition: a comprehensive survey. IEEE Transactions on Pattern Analysis and Machine Intelligence, 22, 63e84. Rayner, K., Chace, K. H., Slattery, T. J., & Ashby, J. (2006). Eye movements as reflections of comprehension processes in reading. Scientific Studies of Reading, 10, 241e255. Rayner, K., & Reichle, E. D. (2010). Models of the reading process. Cognitive Science, 1, 787e799. Reichle, E. D., Rayner, K., & Pollatsek, A. (2003). The EZ reader model of eyemovement control in reading: comparisons to other models. Behavioural and Brain Sciences, 26, 445e476. Reichle, E. D., Warren, T., & McConnell, K. (2009). Using E-Z reader to model the effects of higher-level language processing on eye movements during reading. Psychonomic Bulletin & Review, 16, 1e21. Richardson, D. C., Dale, R., & Kirkham, N. Z. (2007). The art of conversation is coordination common ground and the coupling of eye movements during dialogue. Psychological Science, 18, 407e413. Ra˛ czaszek-Leonardi, J. (2009). Symbols as constraints: the structuring role of dynamics and self-organization in natural language. Pragmatics & Cognition, 17, 653e676. Van Petten, C., & Kutas, M. (1991). Influences of semantic and syntactic context on open-and closed-class words. Memory & Cognition, 19, 95e112. Van Orden, G. C., & Kloos, H. (2005). The question of phonology and reading. In M. J. Snowling, & C. Hulme (Eds.), The science of reading: A handbook (pp. 61e78). Hoboken, NJ: Blackwell. Van Orden, G. C., Pennington, B. F., & Stone, G. O. (2001). What do double dissociations prove? Cognitive Science, 25, 111e172. Wallot, S. (2014). From “cracking the orthographic code” to “playing with language”: toward a usage-based foundation of the reading process. Frontiers in Psychology, 5. http://dx.doi.org/10.3389/fpsyg.2014.00891. Wallot, S., Hollis, G., & van Rooij, M. (2013). Connected text reading and differences in text reading fluency in adult readers. PLoS One, 8, e71914. Wallot, S., O'Brien, B. A., Coey, C. A., & Kelty-Stephen, D. (2015). Power-law fluctuations in eye movements predict text comprehension during connected text reading. In Proceedings of the 37th annual conference of the cognitive science society. Wallot, S., O'Brien, B. A., Haussmann, A., Kloos, A., & Lyby, M. S. (2014). The role of reading time complexity and reading speed in text comprehension. Journal of
Please cite this article in press as: Wallot, S., Understanding reading as a form of language-use: A language game hypothesis, New Ideas in Psychology (2015), http://dx.doi.org/10.1016/j.newideapsych.2015.07.006
8
S. Wallot / New Ideas in Psychology xxx (2015) 1e8
Experimental Psychology: Learning, Memory, and Cognition, 40, 1745e1765. Wallot, S., & Van Orden, G. (2011). Toward a lifespan metric of reading fluency. International Journal of Bifurcation and Chaos, 21, 1173e1192. Webber, C. L., & Zbilut, J. P. (1994). Dynamical assessment of physiological systems
and states using recurrence plot strategies. Journal of Applied Physiology, 76, 965e973. Wittgenstein, L. (1953/2010). Philosophical investigations. Hoboken, NJ: WileyBlackwell.
Please cite this article in press as: Wallot, S., Understanding reading as a form of language-use: A language game hypothesis, New Ideas in Psychology (2015), http://dx.doi.org/10.1016/j.newideapsych.2015.07.006