Semiotica 174 (2009), 49-68
We distinguish between concepts of semiosis in an individual mind and in a community of minds. We show that semiosis, as it occurs in Roentgen diagnosis, differs fundamentally from the Peircean concept of semiosis. At all stages of the argument, basic concepts are derived from the Peircean categories of thought.
Key Words: diagnosis; semiosis; mind; model; complex system; Peircean categories.
In this study, we shall construct a model of semiosis in a community of minds by generalization of a model of semiosis in an individual mind. We first define what we mean by 'mind'. This will entail an examination of the biological substrate of mind or the central nervous system modeled as a type of communication network. At every stage of the argument, basic concepts are derived from the universal categories of thought postulated by Charles S. Peirce (Peirce 1998[1903]):148). We observe that the Peircean concept of semiosis does not apply to Roentgen diagnosis and we propose a concept of diagnostic semiosis that is sufficiently general to be applicable to visual diagnosis in any context. In the first section, the fundamental concept of modeling is reviewed from a categorical perspective, since categorical modeling will be a recurrent theme in what follows.
Models and modeling systems
We have used concepts from the modeling theory of Sebeok and Danesi, (2000) to characterize semiosis in individuals and in communities. In general, models are devices used by inquirers to facilitate inquiry. These devices are representations that draw the attention of an interpreter to selected attributes of objects of inquiry with the exclusion of all others. Such representations involve relations of similarity between attributes of the model and corresponding attributes of its referent. Models, like the representamina of signs, take the place of or 'stand for' their referents in the process of inquiry when the referents themselves are not directly accessible by perception. Hence, the concept of 'model' is an extension of the concept of 'sign'. Both are triadic relations by which a representamen represents a referent in a way that is specified by an interpretant. By convention, the representamen of a triadic sign is commonly referred to as a 'sign' when its interpretant is not specifically an object of inquiry. In the same way, it is convenient to refer to the representamen of a triadic model simply as a 'model' when its interpretant is not itself an object of inquiry. We shall use this convention in the discussion that follows, without losing sight of the fact that any model entails an interpretant. From a pragmatic perspective, models have effects on the mind of the interpreter that facilitate inquiry. In the same way as signs, models may evoke feelings, provoke judgments or impart knowledge in the mind of the interpreter (Cantor 2006).
The pragmatics of modeling
Models as products of mental activity may be externalized or internalized (Sebeok and Danesi 2000:2). The ground of an externalized model is the physical world and the ground of an internalized model is a mind. Clearly, an externalized model may be internalized and an internalized model may be externalized. For an externalized model, there is a similarity of appearance, structure or function between model and referent. This relation is independent of spatial or temporal scale (Black 1962). Internalized models may be classified as architypal, theoretical or mathematical (Black 1962). There is a similarity of form between an internalized model and its referent. An example of an architypal model is the Peircean categories, which constitute a paradigm of the most elementary modes of thought or of 'being' that is accessible by thought. These categories are intuitively expressed by the triad of terms: Attribution, Opposition and Mediation (Firstness, Secondness and Thirdness). Theoretical models may involve analogies between ideas in different conceptual domains or may involve thought experiments (Black 1962). There are many types of mathematical models, since mathematics is the rational study of abstract form. Mathematical models that are pertinent to the study of mind include orderings, lattices and networks (cf. Combs et. al 1958; Harary et. al. 1965).
Modeling systems
Sebeok and Danesi (2000:4) distinguish three types of model by their relation to signs. A singularized model consists of the representamen of a single triadic sign. A composite model consists of contiguous representamina of different triadic signs e.g. a visual scene or a text (cf. Cantor 2003, on compound Roentgen signs). A generalized model consists of shared attributes of referents from different triadic signs. If the referents belong to the same cognitive domain, the model may be termed a code or a generalization. If the referents belong to different cognitive domains, the model may be termed a metaphor or an analogy.
Categorical modeling systems
A modeling system is a device for the production of models. Sebeok and Danesi (2000:10) propose a triadic typology for semiotic modeling systems, based on the Peircean categories:
Inductive logic is a tertiary modeling system. In this system, scientific explication is a modeling process. In this process, a less exact concept is transformed into a more exact concept (Carnap 1962:5). Carnap's criteria of adequacy for scientific explication may be generalized to serve as criteria of suitability and utility for explanatory models (cf. Carnap 1962:7). In this way, criteria of suitability are clarity, fidelity and generality and criteria of utility are simplicity, versatility and uberty.
A pragmatic typology of modeling systems
Three types of modeling systems may be distinguished on the basis of their use. A heuristic modeling system is used as a guide in the process of discovery or innovation. The universal categories constitute the most fundamental heuristic modeling system. A descriptive modeling system is used to represent objects of thought. Language is a descriptive modeling system. Since language is used to represent the universal categories, it is also a heuristic modeling system. A predictive modeling system is used to identify events before they occur. Mathematics and inductive logic are predictive modeling systems. Since these systems are used to represent linguistic descriptions, they are also descriptive and heuristic modeling systems. Hence, this pragmatic typology of modeling systems entails the categorical inclusion rule (cf. Liszka 1996:46).
In this section, we present a brief survey of the models and modeling systems that are available for the rational study of relations. Subsequently, we shall attempt to justify the hypothesis that the biological substrates of personal mind and transpersonal mind may be represented by polyadic functional relations.
Polyadic relations
In this discussion, awareness of a term will be thought of as a first category concept. A dyadic relation is an awareness of two terms in opposition, which is a second category concept. A triadic relation is awareness of mediation between the two terms in a dyadic relation, which is a third category concept. In deductive logic, a binary relation is understood to be an ordered pair of terms (Tarski 1994:82). From a Peircean perspective, binary relations are in fact triadic relations, since ordering is a mediating principle. In mathematics, a relation on a set of element is understood to be a set of distinct binary relations between those elements (Harary et al 1965:6). Hence, a relation in the mathematical sense is reducible to a set of triadic relations. This is an alternative formulation of a 'Reduction Thesis' for relations (cf. Burch 1997). In what follows, we describe tertiary modeling systems that are frequently used to represent polyadic relations.
Graph theory
Graph theory is a tertiary modeling system that is used for the abstract study of polyadic relations or structures (Harary 1965). Since the term 'graph' is used in many different contexts, we shall use the term 'relational graph' in this discussion. Intuitively, a relational graph may be thought of as a set of points that are pairwise connected by a set of lines. Hence, a relational graph is a representation of a set of binary relations. As such, it is a model of a set of triadic relations. The number of points in a relational graph is termed its order. A relational graph is said to be directed if every one of its lines is directed. For computational efficacy, it is assumed that there are no self-connections at any point and no parallel connections between adjacent points. By a further abstraction, a directed graph may be represented by a square array of numbers or matrix. If the points of a directed graph of order n are indexed in an arbitrary way by the natural numbers from 1 to n, one may construct a square array of binary digits consisting of n rows and n columns where the entry in a row i and column j is 1 if the points i and j are adjacent and 0 if they are not adjacent. Since it has been assumed that there are no parallel connections between adjacent points, the adjacency matrix is well defined, once an ordering of points has been chosen. Similarly, a reachability matrix records whether or not there is at least one path of any length connecting points i and j and a distance matrix records the length of the shortest path connecting points i and j. Hence, graph theory is a modeling system for polyadic relations and matrix theory is a computationally efficient modeling system for directed graphs.
Network theory
Network theory is a tertiary modeling system that may be used for the study of changing polyadic relations or functional relations. A network is defined as a directed graph with a constant or variable value assigned to each line. In this way, a network may represent the strength of a changing polyadic relation. Network strength is represented by a value matrix in the same way that connectedness is represented by a reachability matrix. We shall see that the activity of dynamic systems may be modeled by dynamic networks. Statistical information theory is used to detect patterns of change in very large matrices that represent complex dynamic networks e.g. neuronal networks of the central nervous system (Seth and Edelman 2004).
Before dealing with the modeling of complex systems, we must define the sense in which we use the terms 'system' and 'complex'. We first distinguish between a structure and a system:
Categories of complexity
Following Amaral and Ottino (2004), systems may be classified as simple, complicated or complex:
It is possible for systems with a small number of parts to have complex dynamics e.g. a forced pendulum or a double pendulum (Amaral and Ottino 2004). Furthermore, complex systems with changing parts as well as changing interactions may evolve. Following Amaral and Ottino (2004), the phenomenology of an evolving complex system includes:
A categorical basis for the concept of an evolving complex system is given by the following argument:
Hence, the categorical properties of evolving complex systems are emergence (Firstness), adaptation (Secondness) and selection (Thirdness). Note that emergence is spontaneous, that adaptation entails emergence and that selection entails both adaptation and emergence by the categorical inclusion rule.
Modeling complex systems
Modeling systems that are commonly used to represent complex systems include nonlinear dynamics, statistical physics and complex network theory (Amaral and Ottino 2004). Recent work in complex network theory has revealed similar network structures in diverse natural and artificial complex systems (Barabási 2002; Dorogotsev and Mendes 2003). A distinctive property of evolving complex systems is the emergence of hierarchically organized and functionally differentiated subsystems (Girvan and Newman 2002; Ravasz and Barabási 2003). Such patterns of organization are characteristic of biological systems and, what is of primary interest in this study, the central nervous system which is the biological substrate of mind (Sporns et al 2004).
The spatial and temporal scales of biological structures and functions range widely, from molecules to cells, from cells to organs, and from organs to organisms. On any level of organization, biological systems may be resolved into functionally differentiated and functionally integrated subsystems or modules (Hartwell et al 1999). In general, systems may be classified on the basis of similarities or differences in modular structure or function (Barabási et al 2004a).This scheme may be resolved into four functional types characterized as redundancy, versatility, specificity and degeneracy (Edelman 1978:59):
It appears that degeneracy is a necessary condition for the evolution of biological systems. (Edelman and Gally 2001).
The cortex is a dynamic system involving the collective activity of approximately 1010
selectively connected neurons. Although the number of connections is far greater than the number of neurons, the minimum number of interneurons that separate any two neurons is approximately 3 (Braitenberg 1998:193; Sporns and Zwi 2004). Neurons are morphologically similar and physiologically autonomous functional units. The body (soma) of a neuron has ramified fiber-like expansions called dendrites and axons, where the axonic ramifications arise from a main axon. Ramón y Cajal maintained that neurons interact by the propagation of excitatory electrochemical signals across specialized connecting structures (synapses) that form at points of axodendritic and axosomatic contact and that signal conduction is directed toward the main axon (Cajal 1995[1909]: 79, 95). Since the time of Cajal, synaptic connections between every conceivable pairwise combination of dendrite, soma and axon have been demonstrated by electron microscopy (Bullock 1959; Schmitt et al 1976). Furthermore, different types of nonsynaptic interneuronal communication have been described which further complicate the task of structural and functional modeling (Jeffreys 1995). However, there is agreement that synapses are important, if not the major mediators of interneuronal communication, whether by excitation or inhibition, and that transsynaptic signal propagation is unidirectional at any given moment. Furthermore, any neuron may receive signals from multiple neurons and may transmit signals to multiple neurons (Braitenberg 1998:96). Models that represent the cerebral cortex as a type of communication network are based on these concepts.
Modeling the cerebral cortex
In brain network theory, the cerebral cortex is modeled as a directed graph or network with nodes (points) representing neurons, links (lines) representing synapses and with the direction of links representing the direction of signal transmission (Sporns 2003). Clearly, mathematical network theory provides greatly simplified models which do not take into account many aspects of real neuronal networks such as the spatial distribution of neurons, the existence of parallel or antiparallel connections between adjacent neurons (Braitenberg 1998: 97-98), the existence of self-connections (Bekkers 2003) or nonsynaptic interneuronal signaling (Jefferys 1995). A different model is provided by the dual network in which nodes represent synapses, links represent neurons and the direction of each link represents the direction of transsynaptic signal propagation at a given moment. In this directed network, each node (synapse) has only two links. This network models the triadic relation between each synapse and its pre- and post synaptic neurons. It also represents a functional triadic relation: pre- and post synaptic signals mediated by a synaptic threshold.
Categorical rules of synapse modification
Synapses are modified by conditions that alter their activation thresholds. The lowering of the activation thresholds of excitatory synapses is referred to as facilitation. Dynamic models of use-dependent synaptic facilitation were first proposed by Hebb (1949:62, 70; cf. Brown et al 1990).
Hebb's first rule. For a single excitatory synapse, Hebb's first rule may be restated as follows:
If the same presynaptic signal repeatedly provokes a postsynaptic signal, then the synaptic threshold for postsynaptic signal production will be lowered (alternatively, the synaptic resistance to excitation will be decreased). Hence, transsynaptic signaling will be facilitated.
To demonstrate the categorical basis for this rule, we adapt our interpretation of the Peircean categories for use in a dynamic context. In this way, Firstness is action, Secondness is reaction and Thirdness is mediation between action and reaction. Hence, repetition of an efficient presynaptic signal corresponds to temporal similarity of action, provocation of a postsynaptic signal corresponds to temporal contiguity of action and reaction and modification of the activation threshold of the synapse corresponds to a change of convention or mediation. From a historical perspective, it is of interest that Hebb's first rule for the facilitation of transsynaptic signaling was anticipated by William James as a law of neural habit for 'elementary brain-processes' (James 1984[1892]:226).
Hebb's second rule. By one estimate, there are between 104 and 105 synapses per neuron in the human cerebral cortex (Braitenberg 1998:190). We shall refer to the aggregate of synapses located on a single neuron as the synaptic field of the neuron. For the synaptic field of a single neuron, Hebb's second rule may be restated as follows:
If multiple concurrent presynaptic signals repeatedly provoke a postsynaptic signal, then the activation threshold of the synaptic field of the neuron will be lowered i.e. fewer subsequent presynaptic signals will be sufficient to provoke a postsynaptic signal.
As in Hebb's first rule, repetition of concurrent presynaptic signals corresponds to temporal similarity of action, provocation of a postsynaptic signal corresponds to temporal contiguity of action and reaction and modification of the effective threshold of the synaptic field corresponds to a change of convention or mediation. Hence, both Hebbian rules entail the three categorical binary relations of Similarity, Contiguity and Conventionality (cf. Cantor 2004). It must be emphasized that Hebb's rules constitute a simplified model of use-dependent synaptic change that does not take into account synapse generation (synaptogenesis) or nonsynaptic (ephaptic) contact between neurons (Fuster 2003:40-44).
Remodeling the cerebral cortex
When a dynamic system is represented by a network, changes in the system are represented by remodeling of the network. In general, remodeling entails reconfiguration and modulation. Reconfiguration involves the addition or deletion of nodes or links while modulation involves the strengthening or weakening of links. We have seen that the cerebral cortex may be represented by a neuronal or a synaptic network. In such models, reconfiguration follows the generation or elimination of neurons and synapses in ontogenesis (cf. Cowan et al 1984; Macklis et al 2006) and modulation is a result of activity-dependent modifications of morphologically distinct excitatory and inhibitory synapses (Peters 1987).
The central nervous system changes structurally and functionally during development of the individual (ontogenesis) and evolution of the species (phylogenesis). By ontogenesis we mean the entire life-span, prenatal and postnatal, of the individual. Gerald M. Edelman has observed that change in the central nervous system during ontogenesis, like change in phylogenesis, is based on Darwinian principles of variation, adaptation and selection (Edelman 1978; 1987). In phylogenesis, variation is a result of random genetic mutations, adaptation results from conformity of genetic changes to environmental changes, and natural selection results from the differential survival of adaptations. In this formulation, genetic variations are spontaneous, adaptation to environmental change entails genetic variation, and natural selection entails adaptation to environmental change. In ontogenesis, variation involves epigenetic processes acting within genetic constraints (Allis et al 2007), adaptation involves modification of individual variations by contingent events in development and experience, and selection involves differential preservation of developmental and experiential adaptations. In this formulation, genetic and epigenetic variations are spontaneous, developmental and experiential adaptations entail both genetic and epigenetic variations, and selection entails both developmental and experiential adaptations. Hence, the structural and functional diversity of the cerebral cortex of individuals is a consequence of contingent developmental and experiential events constrained by genetic variations (Edelman 1987). Since the categories of complexity are emergence, adaptation and selection, we may consider the cerebral cortex as an evolving complex system.
The cerebral cortex is an organ with mind as its function. In other words, a mind is the ongoing signaling activity of an evolving complex neuronal network in a living brain. Hence, a mind is an ongoing event or a process and not an object. It is an event that occurs within a specific material substrate that occupies a unique spatio-temporal region i.e. a brain. The activity of a unique mind is physically represented by the behavior of a unique body. When all brain activity ceases, the mind and its bodily representations cease to exist.
In his theory of neuronal group selection, Gerald Edelman (1978, 1989) has done much to clarify the biological substrate of mind. In his model, experiential selection is assumed to operate on structurally diverse but functionally similar (degenerate) subnetworks called neuronal groups that are interconnected to form group repertoires. A group repertoire may be thought of as a field of neuronal groups that is continually remodeled by experiential selection rules in the same way that the synaptic field of a neuron is continually remodeled by Hebbian facilitation rules. In ontogenesis, experiential selection is a process of differential adaptation and preservation that acts on fields of functional units (synapses or neuronal groups). In phylogenesis, natural selection is a process of differential adaptation and survival that acts on populations of organisms. Edelman postulates that the activities of different group repertoires are integrated by reentrant signaling, which is a process of continual phasic signaling sustained by parallel and reciprocal connectivity of the underlying neuronal network. This theoretical model has itself been modeled as a physical dynamic system or neural network (a selective recognition automaton) providing evidence for the self-consistency of the theory. In these models, mind is the activity of a type of evolving complex network, characterized as a network of modules with degenerate functionality and reentrant connectivity. It follows that the uniqueness of an individual mind is due to the uniqueness of its biological substrate which is continually remodeled during the life of the individual by developmental and experiential selection (cf. Edelman 2004:141). Furthermore, the privacy of individual mental experience or awareness is a necessary consequence of the fact that an individual mind is the ongoing activity of a specific complex neuronal network i.e. a specific cerebral cortex and its annexes. This precludes a direct awareness of another mind which is the activity of another unique neuronal network. All awareness of other minds is indirect and mediated by signs or models.
In what follows, we adopt the conventional distinction between the terms 'field' and 'domain' due to Csikszentmihalyi (1978). Just as the activity of a unique brain determines a personal mind, the activity of a field of inquirers (a field of minds) engaged in a common domain of inquiry determines a transpersonal mind.
Transpersonal mind as a complex system
A transpersonal mind has mind-like properties that include collective perception, attention, memory and thought. Transpersonal perception is a function of personal minds that communicate their sensory experience to one another through the medium of language. Transpersonal attention is the emergence of the coordinated attention of multiple personal minds directed toward specific aspects of the domain of inquiry e.g. fashions of inquiry or paradigms of thought. Transpersonal memory consists of the communicable memory of personal minds, augmented by specialized instruments for the storage and retrieval of recorded thought e.g. books, libraries, the World Wide Web. Transpersonal thought involves the transfer of thoughts between personal minds by means of language. Clearly, transpersonal mind has a modular structure. While the biological substrate of a personal mind is an evolving network of neurons that interact by electrochemical signals, i.e. a central nervous system, we assume that the biological substrate of a transpersonal mind is an evolving network of central nervous systems that interact by communication. Furthermore, transpersonal mind is a complex process since it changes continually due to:
Modeling transpersonal mind
A field of inquirers having a common domain of inquiry is a type of social network that we refer to as a collegial network. Social network analysis is the formal study of networks of relations between social agents, either individuals or groups (Breiger 2004). Hence, dynamic social network analysis is an appropriate tertiary modeling system for a transpersonal mind, which is defined to be the collective activity of a community of minds. Statistical techniques are used to elucidate the structures and functions of large social networks determined by interpersonal or intergroup interactions such as acquaintance, collaboration, communication, transportation and exchange (cf. Stanley and Havlin 2003). Specifically, collegial networks of scientific collaboration or coauthorship have been investigated by Newman (2001). In the following section, we propose dynamic models of semiosis in both personal and transpersonal minds.
Following Peirce, we define a sign to be a triadic relation in which an opposition between a representamen and its object is mediated by an interpretant. We consider a relation to be a concept in a mind rather than an object in the world. We assume that a relate of a triadic sign may be either a physical or mental object or event.
Relate grounds
When the representamen of a sign is a percept or perceivable object or event in the physical world, the ground of the representamen is in the physical world. When the representamen of a sign is a perception of a physical object or event, the ground of the representamen is the mind of the interpreter. When the object of a sign is a physical object or event that determines the representamen as a perceivable object or event, the ground of the object is in the physical world. When the object of a sign is a memory of a perception or of a representation of a perception, the ground of the object is the mind of an interpreter. We assume that the interpretant ground of a sign is the mind of an interpreter.
Phenomenology of signs
We shall distinguish between phenomenal and mental signs. In a phenomenal sign, the representamen ground is in the physical world and the mind of the interpreter is a common ground for its other relates. In a mental sign, the mind of the interpreter is a common ground for all of its relates. As an object or event in the physical world, the representamen of a phenomenal sign may be transient or persistent. A transient auditory or visual representamen determines a transient phenomenal sign. Transient signs present in the passing scene of perceptual experience. An auditory or visual representamen that is persistent or fixed in its ground determines a persistent phenomenal sign. Roentgen signs represent anatomic events not accessible by direct visualization that are recorded on Roentgen images as persistent signs.
Signs change as their relates change. In general terms, semiosis or the triadic 'action of signs' is a process by which signs change as they are reinterpreted (Peirce 1998[1907]: 411). Peirce conceived of semiosis as an unlimited iterative process of reinterpretation that is hypothetically convergent, in analogy with the mathematical concept of convergence for infinite sequences. Peirce maintained that the interpretant of a triadic sign at one stage in the process of semiosis becomes a representamen in the next stage where it evokes a new interpretant in a new triadic relation with the original object (Peirce 1998[1903]:272-273). However, this model of semiosis does not faithfully represent the real process of sign change by reinterpretation that we encounter in Roentgen diagnosis. Our aim here is to adapt the concept of semiosis to the context of Roentgen diagnosis and to specify a context in which Peircean semiosis does operate. In what follows, we distinguish two types of semiosis that occur in Roentgen diagnosis, termed Peircean semiosis and diagnostic semiosis. Since semiosis involves the mental functions of interpretation and reinterpretation, we shall characterize semiosis in both personal and transpersonal minds.
Semiosis in personal mind
Representamina of the phenomenal signs encountered in Roentgen diagnosis may be transient or persistent in their ground which is a physical image. Such images may be dynamic as in fluoroscopy or static as in radiography.
Peircean semiosis in personal mind. Reinterpretation of a transient phenomenal sign produces a mental sign whose representamen is a memory of the previous interpretant, a memory that evokes a new interpretant and a new triadic relation with the initial object-memory. Subsequent reinterpretations transform mental signs into other mental signs. Hence, transient phenomenal signs initiate a process of Peircean semiosis. Note that this process begins with a change of representamen ground from the physical world to a personal mind. Following this initial change of relate ground, all subsequent signs in the process are mental signs. In this model, it is assumed that visual memory, which is essential for recognition, is not involved in the initial reinterpretation where the previous interpretant serves as a representamen.
Diagnostic semiosis in personal mind. Most of the signs encountered in Roentgen diagnosis are persistent phenomenal signs. Since the representamen of a persistent sign is invariant in its ground (a Roentgen image), reinterpretation of the opposition between the same representamen and the same object-memory results in a new phenomenal sign. This new sign is distinguished by its interpretant, which is based on new collateral knowledge or a new inference. In principle, this iterative process is unlimited. There are obvious advantages in the use of persistent signs in Roentgen diagnosis. Persistence eliminates temporal constraints on the process of interpretation and facilitates consultation between different interpreters. Note that the sign at each stage of the process is a sinsign determined by a specific patient at a specific time.
Semiosis in transpersonal mind
In previous sections, we developed a biologically based model of transpersonal mind, defined as the collective mind-like activity of a community of minds. We have referred to a community of minds with a common domain of inquiry as a field of inquirers. We have seen that communities of common interest or action may be modeled as evolving complex networks. As such, fields of inquirers are not confined to a specific geographic region or a specific period of time (cf. Pallas 2007). New inquirers join a field after training and subsequent accreditation by 'gatekeepers'. Established inquirers may leave the field by loss of interest, retirement or death. In complex networks, a few individuals become 'hubs' with exceptionally high connectivity and certain subnetworks become 'hot spots' of exceptionally high activity (cf. Barabási 2002; Barabási et al 2004b). We have seen that communities of minds have mind-like attributes including collective modes of perception, attention, memory and thought. In this section, we characterize semiosis in the transpersonal mind of a field of inquirers engaged in a specific domain of inquiry: medical diagnosis by the interpretation of Roentgen images.
Peircean semiosis in transpersonal mind. It has been noted (Cantor 2004) that types of inference used in Roentgen diagnosis differ from the types of inference that are acceptable in scientific inquiry. However, both Roentgen diagnosis and scientific inquiry employ empirical methods that have the potential to produce a provisional consensus i.e. a consensus in a field of inquirers that is subject to revision. Provisional consensus may only be achieved by reference to a standard that is independent of mind i.e. the physical world. We find that Peircean semiosis is an inappropriate model for the reinterpretation of phenomenal signs by a transpersonal mind. If the interpretant at one stage of the process becomes a representamen in the next stage, a phenomenal sign will be transformed into a mental sign that will fail to satisfy pre-established criteria of representational adequacy. This change will preclude the correction of interpretive deficiencies or errors and exclude the possibility of achieving a provisional consensus by 'convergence' in any reasonable sense of the word. However, we do believe that non-convergent semiosis is an adequate model for the semiosis of mental signs e.g. in spoken language.
Diagnostic semiosis in transpersonal mind. A phenomenal sign may be reinterpreted in transpersonal mind if its representamen is transient but reproducible in its ground or if it is persistent in its ground. Such signs may be reinterpreted by different personal minds at different times in the process of diagnostic semiosis. The representamen at each stage of this process is an exemplar of a legisign retrieved from storage as an image in transpersonal memory. In this way, diagnostic semiosis is a model for semiosis in transpersonal mind.
Signs for which both the representamen ground and the object ground are the physical world are termed material signs. The interpretant ground of a material sign may be a personal mind or a transpersonal mind. The relates of a material sign interact in various ways:
Hence, material signs are triadic relations with interacting relates i.e. triadic systems. For material signs with transpersonal mind as their interpretant ground, diagnostic semiosis is an unlimited process of interpretant change (Figure 1). Since interpretant change is emergent within mind, adaptive to memory and selective by criteria of adequacy, semiosis is a process by which signs evolve. Hence, material signs are evolving systems. Furthermore, semiosis of material signs within a transpersonal mind and constrained by the physical world may be considered a directed or 'purposeful' process. This teleology is derived from the intentionality induced in a transpersonal mind by its constituent personal minds.
We have conceptualized models for diagnostic semiosis in both personal and transpersonal minds. Although these models were devised to represent semiosis as it occurs in Roentgen diagnosis, they are clearly applicable to any type of visual diagnosis. We have seen that semiosis is a process of mind where mind is a function that emerges at the highest levels in a hierarchy of evolving complex systems modeled as networks. We have shown that such conceptual models may be derived from the Peircean categories of thought.