Our machine builds (or represents) its world, including other machines, within a hierarchy of abstractions, in order to establish the analogic connections that will allow it to identify (self-induct) itself with an other by self-
inducting
itself into the future not as itself, but as the not-I represented in A/O LTM.
Brett Bourbon - 1996 - Constructing a Replacement for the Soul
Further reproduction prohibited without permission.
599
? to retest the match, which will have to match not only in value but also in the O memory slot it is found. This means that a test will take at least as long as 5 moments, slow but necessary if a district is as wide as the distance it takes sound to travel for 5 machine moments(itmightbefilledwithsomedensegasorevenwater). Thisretestingwillalso reduce the problem o f false matches (where a degree o f change value o f +2 could be obtained by a 5 moment followed by a 3 moment and a 3 moment followed by a 1 moment). It is, o f course, not even necessary that a real match, from our perspective, take place. Itisenoughthatitseemslikeamatchtothemachines(althoughanyfalsematch willdissolvefairlyquickly. Itisunlikelygiventhenumberofpossiblecombinationsinour 5 object system that the difference ratio between moments would remain the same in two
different districts for more than 3 moments). The machine edits out A inputs different from its own internal state as "noise" once a match has been discovered.
14. 11 Looking for Mr. Goodmachine
If a machine fails to find an auditory match with its own O Difference Detector
after 3 vocalizations ( in order to give machines which have not yet vocalized but are within the same district time to communicate with this other machine before it moves), it moves the distance possible in 3 moments, and then trumpets its degree o f change value. (The machine has eyes on the sides o f its head so that it can see, and thus function within theprimaltime-scalethatdeterminesitsmomentsanditsrateofvocalization). Ifit processes a match in transit, the machine will stop and retest the match. If it finds a match then it doesn't move. Both machines are in the same district and, therefore, happiness!
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
600
? Once a positive match has been made three things will happen. First, a pattern detector will correlate a signal from the Optical Difference Detector sent every Optical moment with its matched auditory input. It will discover that the match happens every 5 moments. Second, the correlated auditory and optical information is stored in Auditory/Optical Long-Term Memory (A/O LTM). Third, a signal is sent from the
A/O mediating agent to V in order to re-initialize the timing for vocalizations to match the vocalizationsofthemachinethatiswithinthedistrict. (Iamignoringthecaseofmore than one machine in the district, although the districts are large enough to accommodate more than one machine). At the Vocalizing Agent a difference detector is installed in order to detect the change from the previous vocalizing timing. This information is then sent to A/O LTM. The A/O LTM, therefore, is organized as a series of bifurcated units representing an Auditory value and an equal Optical value within the domain marked by the change in vocalization timing.
14. 12 Meta-temporal identities: the modeling of others as syntax
The A/O Evaluator has constructed a meta-temporal identity within LTM, an
identity (not yet an entity) that organizes temporal information (two levels of abstraction above the initial perception of change), instead of representing particular temporal states. The A/O Evaluator has integrated the O agent and A agent, and in so doing has created a more abstract and mediated information structure. The information in the A/O LTM, because it is distinguished from the machine's own vocalizations, represents the external world. This is an external world, however, that includes representations of what the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
601
? machine recognizes as internal states. By matching A input with its own phenomenal perception the machine models itself within itself by modeling another machine (although the machine understands this as simply a sensory input). As a model for humans, this describes consciousness as a means ofconstructing a model ofthe minds ofother humans in order to predict their actions and manipulate them.
It is not quite so simple, however, to ascribe this A input to another machine when itisimpossibleforonemachinetoidentifyanotherasthesourceofthisinput. The machine must first recognize its internal state, the information generated by A, V, and O, as its own internal state. In my machine self-realization emerges from the synchronization o f the machine's vocalization with the vocalization represented by the matched auditory input. This synchronization is a kind of internal state change, caused, however, by an external stimulus. Because the machine is able to separate out its own vocalization, it can distinguish between I and not-I in relation to this vocalization and its recognition as a vocalization generated from its optical system. By matching the auditory input with its O Evaluator input it constructs an informational object that represents the not-I as the I, and the I as a not-I. This information, the meta-temporal identity, can have not meaningful significance, however, before it is used in some control function.
The short-term memories generated by those auditory inputs that do not match the machine's own degree o f change value have little significance, until the matched inputs cease. Once this happens, there is no longer an internal representation o f the temporal history o f the A agent. It is critical for our objective o f constructing a temporal syntax o f thefuturethatthisnothappen. Whilesuchacontinuoushistorydoesnottakeplaceinour
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
602
? brains, an analogous history is written in our external representations of time. The earliest o f these are the notations and sticks and bones, identified by Marshack, marking the lunar cycle. There are curious parallels between these 35,000 year old markings and the internal symbolic economy ofthe machine. These markings are a record ofexperimental observations. One cannot imagine a use for them unless their exists a cultural logic, supported by long-term prediction and planning, into which the notations fit. Once could imagine something like, "Every other lunar cycle, at the first quarter moon and the final new moon, we will do X. " These markings become signals both of time and correlated human activity. The need to determine events through this kind of record keeping suggests a complicated social organization that requires some objective form of organizing its activities, and which supports a wide variety of cultural activities (where questions of correct timing might cause conflict without an "objective" time map).
The markings, however, are non-arithmetic. This means the lunar cycle is not defined by a number o f days (nor are any o f the phases), but by grouping markings in a general pattern. Thus the number of days may vary between the full and half moon in any month (depending on the observations of the scribe), but their basic temporal relationship remainsthesame. ThiskindofgroupingissimilartothecreationofdomainswithinA/O LTM. Thus the regularity of the moon changes is analogous to the less regular, but still periodic, vocal timing changes. In the human mind the creation of organizing syntactical domains is largely a social phenomena, and thus needs to be externalized. Because my machines only form very simple social groups, these structures cannot be supported by culture.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
? The ability to perform this kind o f representational abstraction requires the same kind o f ability to derive form from particular content that I am trying to build into the Time Machine. For early humans, a single mark stands for the idea, the concept of a unit relative to the lunar cylce, and not its particular content (except to the degree that an event or action is meaningful in relation to the logic of the notational series as a whole).
Similarly the grouping o f these units, because they lose the particularity o f their content, can be used to "mean" a particular lunar period. These kinds of representational systems are symbolic languages, where markings, like words, are generalized in their meanings (word less so than the marks for days) so that they can be exchanged and organized within ameta-structurenotoverwhelmedbydetail. Symbolizationisaprocessofsimplification that allows an event to be reduced to a particular representation that can then be used to form other meaningful statements. All units within these lunar calendars must be relatively identical, in the way that words define events in relation to pre-established categories of meaning.
For our machine a continuous representation of time will be used, like it was in early human bone markings, to define a continuous temporal structure in which all machine activity can take place. If the A input ceases to match the O Evaluator values, the A unmatched STM is shunted and nested within the A/O LTM. By nesting I mean the informational structure generated in the A unmatched STM is translated into the language, the form, of the A/O LTM. The A unmatched STM provides a temporal link betweentheendingofonemachinematchtothebeginningofanother. Theinformationit stores in A/O LTM must fit within the parameters previously set up to accommodate the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
604
? matched information. The unmatched information (e. g. a difference o f -3 between the machines internal Optical difference and its auditory input) has no meaning in relation to thematcheddifferencevaluesof0,unlesssimplytoindicatematchorunmatched. This kind o f simplification, or abstraction, is necessary if the unmatched STM is to function in relationtoA/OLTM. Theeffectofsuchanesting,therefore,istoignorethecontentof the Auditory (A) unmatched STM, and simply use this information to mark the primary temporal units constructed in the Optical Agent. The structure of A unmatched STM is used to augment and continue the structure ofAJO LTM. . Once this is accomplished, the A/O LTM can be understood as a continuous, simplified representation, through external surrogates,oftheinternal temporalstructureofthemachine.
Because the A input and O input values are correlated with each other, their content is in effect erased. Because each memory slot in A/O LTM encodes so little information, it becomes in effect a series o f marks which are themselves marked by the change in vocal timing. Sensory input is organized within a structure generated in relation to an internal state change. The A/O LTM models both external and internal time within asinglestructure. Evenmoresignificantlythelevelofabstractionoftheinformationused by the A/O Evaluator has allowed the creation of a model of the structure of time divorced from its content (defined in relation to a unit o f difference, not an instant in time). Our problem now is to figure out how to use this structure.
14. 13 Negative entropy
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
605
? The A/O LTM gives our machine the ability to make a long range prediction. We must assume that at least one machine can outlive a number of its matches. If so, the only pattern a memory analyzer o f A/O LTM could identify would be that every vocal synchronization change will be followed by another change after a period o f time. Because the content of the A/O LTM is so poor, if a change in vocal synchrony occurs and recalls the memory of past changes (that is, the pattern of such changes), it will recall the temporal structure embodied in the LTM. Thus, whenever a new match occurs a long-range prediction would be possible: this vocal change will be followed by a series of differences (temporal units) which will themselves end in a new vocal timing change.
The structure supporting this prediction (the moments represented by the O Difference
Detector) will be in place until this expectation (the next vocal timing change) has been met, at which point it will begin again. Attached to the A/O Evaluator will be a memory recall agent that will work in parallel with the Evaluator, but will also organize both A and O input within the temporal structure imported from the A/O LTM. We have a future.
14. 14 Re-building ourselves in the other
This leaves us with two problems however. How can we integrate the other
perceptual, evaluatory, and memory agents into this structure? And what is the significance o f the two states (match and unmatched) within the temporal series in the A/O LTM?
The notion of the future, dependent as it is on a society of machines (identical in everything but life-span), requires a society of different sensory agents. Is it necessary,
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
606
? however, to integrate the O LTM, with its patterns and short-term predictions, within the long-term future of A/O agent? If our machine receives optical information that triggers a short-term prediction, we might want this to signal the A/O Evaluator. But the A/O agent could not understand this kind of pattern (between numbers of objects). Such a pattern would not be reflected in its degree of change information. The only patterns it could recognize, other than the change in vocal synchronization, is the shift from its shunting of matched to unmatched information to LTM and a correlation between a matched auditory inputandeveryfivesignalsfromtheOpticalDifferenceDetector. IftheAagentcould activate a pattern recognition agent to evaluate the unmatched A STM that would be sensitive to more abstract patterns o f difference (every degree o f change 2 is followed by a degree o f change 4), and if we pretend that these patterns are present in some districts, then our two memory structures could be integrated easily by running them both through
the same memory recall agent. The existence o f such patterns in the machine world would indeed be convenient, but then so is the bipolar structure of water.
We have three sub-agents, constituting three internal time scales, not fully exploited in modeling time: the (1)V Difference Detector indicating a change in vocal timing, the (2) A/O Evaluator's awareness that every match happens after 5 optical moments, and the (3) A/O Evaluator's (one o f its internal agents) ability to detect the shift from matched to unmatched outputs to memory. (1) and (3) constitute long-term regularities, while (2) is a short-term prediction that can act as the structural content organizing the temporal structure enabling long-term predictions.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
607
? The machine's change in (1) vocal timing finds its analogue in the death o f the external matched signal: the first registers as an internal change; the second as an external change. We still do not know if our machine understands this as the death o f its machine mate. An existential understanding ofthe loss ofthe matching voice would require (ifwe ignore the lack o f a mammalian brain, or some might say any brain) an analogic equivalence between the A/O Evaluator shift to unmatched memory output and the death o f itself. It seems unlikely that a machine without a more developed symbolic life could
understand this loss. The predictive power of this kind of system, however, remains stimulus dependent. When the stimulus vanishes, the future disappears.
14. 15 Thepursuitofdeath
If we could make a higher level analyzing agent to determine and chart the larger patterns of change in A/O LTM [(3): matched to unmatched], we might be able to stimulate the machine into a weak awareness of the subjectivity of another machine encoded in the A/O LTMs meta-temporal identity. The A/O Evaluator sends a signal to
this agent when the shift from matched to unmatched occurs, and this signal causes this Pattern Recognizer to recall previous shifts and their defining vocal timing domains. The Pattern Recognizer could construct an information structure that charts the internal state changes against the representation o f catastrophic loss o f the external matching voice.
The information such a chart could encode is rather simple, and the prediction that follows even simpler: every vocal timing change will be followed not only by another, but also by the loss ofthe not-I representation. Because the representation ofthe not-I is used to
with permission of the copyright owner. Further reproduction prohibited without permission.
? determine the actions o f the machine (it stays and observers), its loss results in a change in action (the machine now moves). Before moving can be a form o f mourning, it must also representalossofbeingforthemachine. Ifweremovethemachine's eyesonthesideof its head, it will lose its primary temporal portal. It will still have its auditory input (and STM) by which it can chart a different temporal order, but the loss of this primary information will be recorded with the shift from matched to unmatched in the A/O LTM and will be noted by the A/O LTM Evaluator. This sensory loss, and the time/ memory scale it constructs, can stand for, can symbolize, the loss o f the external voice and the change in vocal timing. The prediction that a loss o f the external voice will necessarily follow a change in the machine's own vocal timing will also be tied to another internal
effect: thecommandtomoveanditssubsequenttemporarylossofsightandthetime- scale that sight generates. This causal structure (Cause [external loss] and Effect [internal loss]) mirrors the vocal timing change and is nested within the basic structure generated from that timing change. Our sense o f the future is growing more complex.
14. 16 Learning from the future
The bifurcation o f the memory and predicative domain defined and signaled by the
vocal timing change sets up a meta-time scale at which level internal symbolic exchanges begin to make some kind o f sense. The O agent wants to be able to predict when it will fail, and when it's short-term predictions will be irrelevant, while the A/O wants to be able to predict when it will have to shift its memory output from matched to unmatched. Our machine could not discover such detailed information, nor could these agents as they now
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
609
? stand actually want anything. They would first have to need this information as a means to act. But what is important here is the conceptual problem this presents. For the Optical Agent the loss of the external voice could stand for its loss o f sight, and because it happens prior to its effect, it could serve as a short-term prediction (generated out of a long-term prediction structure). This represents an internal symbolic communication, where one loss can stand for another. The signal itselfwould go from the A/O Evaluator to the O Agent. This signal is not, however, caused by an algorithm: if shift to unmatched output, send a signal to O that loss of sight is imminent. The meaning ofthe shifttounmatchedoutputisdeterminedbytheA/OLTMPattern. Itconcludesfrom previous patterns that the shift means a loss o f sight. This is a conclusion based on an internal analysis (abstraction) from a lower level abstraction (A/O LTM) which is in turn
generated from lower level sensory modeling. This is not the most elegant of control structures (but neither is much of our learning compared to "lower" animals: human weavingcomparedwithspiderweaving). Ithastheadvantage,however,ofmakingall predictions a function ofmemory (they are learned), and not of pre-programming.
14. 17 Inside a Chinese box: when structure becomes content within larger structures
A matched memory state (any input of value 0) within the A/O LTM, re- interpreted in terms o f the time scale generated from the Optical Agent, stands for 5 Optical moments. If a match has been found, then every five O moments will result in a matched input to A/O LTM. The Pattern Recognizer attached to A/O Evaluator
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
610
? functions as a symbolic catalyst linking the two time scales. The Pattern Recognizer has the power to include in any memory state within A/O LTM a representation of the 5 O moments. (Unfortunately,ourmachineistoosimpletoneedthiskindofsymbolic transfer. Without this need and use all one can say is this machine has the conceptual structure that could allow this integration o f two time-scales as a form o f its temporal consciousness).
14. 18 The I as the not-I: using the future to build other machines
Things are not so easy for the A/O LTM Pattern Recognizer (not to be confused
with the A/O Evaluator's Pattern Recognizer). It can learn to expect the loss of the matched input, but not when, except to the degree that it can know it will happen before the next vocal timing change and before the next loss of sight. In fact the initializing vocal change and the loss of sight define the domain of the matched memory. At the level of abstraction found in the A/O LTM Pattern Recognizer this is enough information to constructourmeta-temporalidentityinexistentialterms: adomaindefinedbytwo internal state changes and consisting of a model of internal perceptual states, but which are
notthoseperceptualstates. Thelossofsightcanstandforthelossofthisdomain. A feedback loop from the A/O LTM Pattern Recognizer could signal the loss of the matched domain to the A/O Evaluator, and thus eliminate the previous hard-wire instruction to shift its output to LTM when it no longer computes a match between A inputandOinput. ThetranslationofauditoryinformationintoalanguagetheOagent can understand ("you will lose sight"), and the reverse translation ("you must shift to
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
611
? unmatched memory"), in spite ofthe physical improbabilities, constructs the symbolic structure that allows for vocal timing changes to stand for a future state o f the machine. Because this state is in the future, it is a not-I. Because this state, while recognized as the representation of the machine's internal state, is built from and also represents an external input, it is a not-I. It is easy for the machine to recognize itself in the auditory external input (the not-I as I). But much more difficult to recognize the I as a not-I. The future provides the "distance", the abstraction that allows the I (the internal state of the machine) to function as a not-I equivalent to the not-I it already recognizes as a representation of itself. Thus, the machine cannot quite recognize the death of the other machine, but it can recognize the loss o f the external auditory signal that it uses to model its own internal state with the new and different life (internal state) the future offers it. It can believe in the possibility of the future.
This is kind of self-deception, used like fiction, to jump into the possibilities that logic cannot capture. It makes the machine what Stanislaw Lem might call a self- inductingmachine. WhenErgtheSelf-Inductorwounduptheclockprincesswiththekey to her life, stolen by a terrible human paleface, he had to play with time (and make up stories) in order to hide the fact that he had simply fabricated a new key instead of pursuing the mad human through space.
Our machine builds (or represents) its world, including other machines, within a hierarchy of abstractions, in order to establish the analogic connections that will allow it to identify (self-induct) itself with an other by self- inducting itself into the future not as itself, but as the not-I represented in A/O LTM.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
? Our machine builds a temporal map in order to make long term predictions. The Optical memory recall Agent can be nested within the A/O LTM recall Agent, thus integrating both time-scales within a single time sense. This is possible because the temporal structure that the A/O LTM Recall Agent remembers is both continually active anddevoidofanycontent. Eachmomentissimplyastructuralmark,and,becauseitis defined by degree o f change values, it recognizes not single phenomenal states, but the dynamic two state moments stored in O STM and O LTM.
In a more complex organism like human beings, however, this future structure, built out of internal changes caused by external inputs that simultaneously represent an other and the self (that is, there is a distinction between other and self, but the other can sometimes be used to represent the self), can abstract others to the point that they can stand for oneselfnot only in the future possibilities of life, but in the future possibilities of death. Outofthisallculture,includingtechnologyandlanguageisbuilt.
1This exercise can seem as a conceptual counterpoint to Heidegger's analytic description of our being towardthefutureinBeingandTime. Iamlookingatdifferentwaysinwhichbeing(orbecoming)canbe described. Iamnottryingtodescribebecomingatallhere. Rathermymachinesaremeanttofunctionas such a description.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
613
? Epilogue: But what have I denied the existence of?
I claimed earlier that my machine(s) would serve as an allegory about how ontological limits describe ways ofmaking sense through the process ofbuilding. In what way are these machine(s) allegories? Commonly, an allegory is an ordered set o f relations withwhichorintowhichatextorsituationiscorrelatedortranslated. Inthissense allegories are applied. Such applications are fraught with dangers and confusions. A possible interpretation is not always probable. Furthermore the application o f an allegory can confuse the interpretive relations within the allegory with a causal explanation of events or histories (e. g. Freud or many political allegories). My machine allegories are very different. In fact in themselves they are not meant as allegories at all. The process of making them, o f correlating aspects o f my experience o f the world in relation to a clear set of ontological limits should be understood as allegorical. I am not writing an allegory of things or o f the world these things inhabit but an allegory o f the making o f these things and their world. The allegory does not have content, but the process of making the machine(s)does. Thisisanallegoricalprocessinwhichtheengineeringlimitsconfronted in building the machines opens the world as a system o f limits, organized around
causation, but directed at removing black boxes (homunculus or God or soul or the unconscious). Suchremovalcreatesanallegoryoflimitswithinwhichwebuildaspectsof sense as a way o f orienting ourselves within the world and or experience.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
? My machines have few places to go, confined as they are to their world, but they haveawhentogointoorremainwithin. Thiswhenisasyntaxwithaverysimple semantics. Their processing builds a temporal envelope whose surface is constructed through the mutual correlation of a machine's own actions (vocalization) with its recognition o f another and with the limits o f the world (not to mention the perversity o f its creator). In many ways the construction o f a similar envelope has been the project o f this dissertation. I have attempted to construct a context within a set of texts in order to make visible the ways in which our negotiations within and towards sense and nonsense generate temporallimits. Howweinhabit'I'and'our'expressesourinvolvementinlanguageat the level o f symbolic functioning that confuses the future and the present (time) and meaningandtheword(grammar). Wefunctionasandwithinfragmentsoftimeand language through which we expose and disguise the limits of how we are, not just what or
who we are.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
615
? Bibliography
Abrams, M. H. The Mirror and the Lamp: Romantic Theory and the Critical Tradition. NY: Norton, 1953.
Adams, Henry. The Education o fHenry Adams. Ed. Ernest Samuels. Boston: Houghton Mifflin Co. , 1918, 1946, 1973.
Adorno, Theodor W. Negative Dialectics. Trans. E. B. Ashton. New York: Continuum P, 1966.
Adorno,Theodor. Quasiunafantasia: EssaysonModemMusic. Trans. Rodney Livingston. NY: Verso, 1992.
Ammons, A. R. Shere: The Form o fMotion. NY: WW Norton, 1974.
Arbman, Ernst. "Untersuchungen zur primitiven Seelenvorstellung mit besonderer
Rucksicht auflndien. " In Le Monde Oriental 20 (1926) 85-222 and 21 (1927) 1- 185.
Aristotle. Aristotle On the Soul. Trans. W. S. Hett. London: Heinmann. Loeb Clasical Library, 1936.
_______. Aristotle's Metaphysics, Books Zeta, Eta, Theta, Iota. Trans. M. Furth. Indianapolis: Hackett, 1982.
_______. The Physics. Trans. R. P. Hardie and R. K. Gaye. In The Basic Works o f Aristotle, ed. Richard McKeon (New York: Random, 1941).
Ashberry, John. Flow Chart. NY: Alfred A. Knopf, 1992.
____________. ReprortedSightings: Art Chronicles 1957-1987. Cambridge: Harvard
UP, 1991.
Atherton, James S. The Books at the Wake: A Study o fLiterary Allusions in James
Joyce's Finnegans Wake, Carbondale: Southern Illinois UP, 1974.
Aquinas, Thomas. Summa Theologia. Latin and trans. Cornelius Ernst O. P. (NY:,
1972); Summa Theologiae: Cura Fratrum eiusdem Ordinis, 5 Vols. (Madrid: Biblioteca de Autores Cristianos, 1994).
Augustine. Confessions. Trans. Henry Chadwick. NY: Oxford UP, 1991.
________. De Trinitate. In Augustine: Later Writings. Ed. John Burnaby. Philadelphia:
The Westminster P, 1955.
Austen, J. L. "A Plea for Excuses". In Philosophical Papers. Oxford: Oxford UP, 1979. Bach, Kent. Thought and Reference. Oxford: Clarendon P, 1987.
Barnes, Jonathan. The Presocratic Philosphers. Boston: Routledge and Kegan Paul, 1979. Barth. Karl. The Epistle to the Romans. Trans. Edwyn C. Hoskyns. Oxford: Oxford UP,
1933.
_________. The Theology o fJohn Calvin. Trans. Geoffrey W. Bromiley. Grand Rapids:
William B. Eerdmans Pub. , 1995.
Beckett, S. et al. OurExagmination RoundHistFactificationfor Incamination of Work
in Progress: James Joyce/Finnegans Wake: A Symposium. NY: New
Directions, 1929.
Benjamin, Wlater. The Origin o f German Tragic Drama. Trans. John Osborne. NY:
Verso, 1977.
______________. "Re the Theory ofKnowledge, Theory ofProgress," In Benjamin:
Philosophy, Aesthetics, History. Ed. Gary Smith. Chicago: U of Chicago P, 1989.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
616
? Berger,Harry. SecondWorldandGreenWorld:StudiesinRenaissanceFiction-Making. Berkeley: U ofCaliforniaP, 1988.
Benveniste, Emile. Indo-European Language and Society. Trans. Elizabeth Palmer. Coral Gables, Fla. : U ofMiami P, 1973.
______________ . Problems in General Linguistics. Trans. Mary Elizabeth Meek. Coral Gables: U ofMiami P, 1971.
Bishop, John. Joyce'sBookoftheDark. Madison: UofWisconisnP, 1986.
Bishop, Elizabeth. The Complete Poems: 1927-1979. NY: Farrar Straus Giroux, 1979. Blake, William. The Poetry and Prose o f William Blake. Ed. David V. Erdman. Garden
City: Doubleday, 1965.
Blanchot, Maurice. " The Athaneum" In The The Infinite Conversation. Trans. Susan
Hanson. Minneapolis: U o f Minnesota P, 1993.
599
? to retest the match, which will have to match not only in value but also in the O memory slot it is found. This means that a test will take at least as long as 5 moments, slow but necessary if a district is as wide as the distance it takes sound to travel for 5 machine moments(itmightbefilledwithsomedensegasorevenwater). Thisretestingwillalso reduce the problem o f false matches (where a degree o f change value o f +2 could be obtained by a 5 moment followed by a 3 moment and a 3 moment followed by a 1 moment). It is, o f course, not even necessary that a real match, from our perspective, take place. Itisenoughthatitseemslikeamatchtothemachines(althoughanyfalsematch willdissolvefairlyquickly. Itisunlikelygiventhenumberofpossiblecombinationsinour 5 object system that the difference ratio between moments would remain the same in two
different districts for more than 3 moments). The machine edits out A inputs different from its own internal state as "noise" once a match has been discovered.
14. 11 Looking for Mr. Goodmachine
If a machine fails to find an auditory match with its own O Difference Detector
after 3 vocalizations ( in order to give machines which have not yet vocalized but are within the same district time to communicate with this other machine before it moves), it moves the distance possible in 3 moments, and then trumpets its degree o f change value. (The machine has eyes on the sides o f its head so that it can see, and thus function within theprimaltime-scalethatdeterminesitsmomentsanditsrateofvocalization). Ifit processes a match in transit, the machine will stop and retest the match. If it finds a match then it doesn't move. Both machines are in the same district and, therefore, happiness!
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
600
? Once a positive match has been made three things will happen. First, a pattern detector will correlate a signal from the Optical Difference Detector sent every Optical moment with its matched auditory input. It will discover that the match happens every 5 moments. Second, the correlated auditory and optical information is stored in Auditory/Optical Long-Term Memory (A/O LTM). Third, a signal is sent from the
A/O mediating agent to V in order to re-initialize the timing for vocalizations to match the vocalizationsofthemachinethatiswithinthedistrict. (Iamignoringthecaseofmore than one machine in the district, although the districts are large enough to accommodate more than one machine). At the Vocalizing Agent a difference detector is installed in order to detect the change from the previous vocalizing timing. This information is then sent to A/O LTM. The A/O LTM, therefore, is organized as a series of bifurcated units representing an Auditory value and an equal Optical value within the domain marked by the change in vocalization timing.
14. 12 Meta-temporal identities: the modeling of others as syntax
The A/O Evaluator has constructed a meta-temporal identity within LTM, an
identity (not yet an entity) that organizes temporal information (two levels of abstraction above the initial perception of change), instead of representing particular temporal states. The A/O Evaluator has integrated the O agent and A agent, and in so doing has created a more abstract and mediated information structure. The information in the A/O LTM, because it is distinguished from the machine's own vocalizations, represents the external world. This is an external world, however, that includes representations of what the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
601
? machine recognizes as internal states. By matching A input with its own phenomenal perception the machine models itself within itself by modeling another machine (although the machine understands this as simply a sensory input). As a model for humans, this describes consciousness as a means ofconstructing a model ofthe minds ofother humans in order to predict their actions and manipulate them.
It is not quite so simple, however, to ascribe this A input to another machine when itisimpossibleforonemachinetoidentifyanotherasthesourceofthisinput. The machine must first recognize its internal state, the information generated by A, V, and O, as its own internal state. In my machine self-realization emerges from the synchronization o f the machine's vocalization with the vocalization represented by the matched auditory input. This synchronization is a kind of internal state change, caused, however, by an external stimulus. Because the machine is able to separate out its own vocalization, it can distinguish between I and not-I in relation to this vocalization and its recognition as a vocalization generated from its optical system. By matching the auditory input with its O Evaluator input it constructs an informational object that represents the not-I as the I, and the I as a not-I. This information, the meta-temporal identity, can have not meaningful significance, however, before it is used in some control function.
The short-term memories generated by those auditory inputs that do not match the machine's own degree o f change value have little significance, until the matched inputs cease. Once this happens, there is no longer an internal representation o f the temporal history o f the A agent. It is critical for our objective o f constructing a temporal syntax o f thefuturethatthisnothappen. Whilesuchacontinuoushistorydoesnottakeplaceinour
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
602
? brains, an analogous history is written in our external representations of time. The earliest o f these are the notations and sticks and bones, identified by Marshack, marking the lunar cycle. There are curious parallels between these 35,000 year old markings and the internal symbolic economy ofthe machine. These markings are a record ofexperimental observations. One cannot imagine a use for them unless their exists a cultural logic, supported by long-term prediction and planning, into which the notations fit. Once could imagine something like, "Every other lunar cycle, at the first quarter moon and the final new moon, we will do X. " These markings become signals both of time and correlated human activity. The need to determine events through this kind of record keeping suggests a complicated social organization that requires some objective form of organizing its activities, and which supports a wide variety of cultural activities (where questions of correct timing might cause conflict without an "objective" time map).
The markings, however, are non-arithmetic. This means the lunar cycle is not defined by a number o f days (nor are any o f the phases), but by grouping markings in a general pattern. Thus the number of days may vary between the full and half moon in any month (depending on the observations of the scribe), but their basic temporal relationship remainsthesame. ThiskindofgroupingissimilartothecreationofdomainswithinA/O LTM. Thus the regularity of the moon changes is analogous to the less regular, but still periodic, vocal timing changes. In the human mind the creation of organizing syntactical domains is largely a social phenomena, and thus needs to be externalized. Because my machines only form very simple social groups, these structures cannot be supported by culture.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
? The ability to perform this kind o f representational abstraction requires the same kind o f ability to derive form from particular content that I am trying to build into the Time Machine. For early humans, a single mark stands for the idea, the concept of a unit relative to the lunar cylce, and not its particular content (except to the degree that an event or action is meaningful in relation to the logic of the notational series as a whole).
Similarly the grouping o f these units, because they lose the particularity o f their content, can be used to "mean" a particular lunar period. These kinds of representational systems are symbolic languages, where markings, like words, are generalized in their meanings (word less so than the marks for days) so that they can be exchanged and organized within ameta-structurenotoverwhelmedbydetail. Symbolizationisaprocessofsimplification that allows an event to be reduced to a particular representation that can then be used to form other meaningful statements. All units within these lunar calendars must be relatively identical, in the way that words define events in relation to pre-established categories of meaning.
For our machine a continuous representation of time will be used, like it was in early human bone markings, to define a continuous temporal structure in which all machine activity can take place. If the A input ceases to match the O Evaluator values, the A unmatched STM is shunted and nested within the A/O LTM. By nesting I mean the informational structure generated in the A unmatched STM is translated into the language, the form, of the A/O LTM. The A unmatched STM provides a temporal link betweentheendingofonemachinematchtothebeginningofanother. Theinformationit stores in A/O LTM must fit within the parameters previously set up to accommodate the
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
604
? matched information. The unmatched information (e. g. a difference o f -3 between the machines internal Optical difference and its auditory input) has no meaning in relation to thematcheddifferencevaluesof0,unlesssimplytoindicatematchorunmatched. This kind o f simplification, or abstraction, is necessary if the unmatched STM is to function in relationtoA/OLTM. Theeffectofsuchanesting,therefore,istoignorethecontentof the Auditory (A) unmatched STM, and simply use this information to mark the primary temporal units constructed in the Optical Agent. The structure of A unmatched STM is used to augment and continue the structure ofAJO LTM. . Once this is accomplished, the A/O LTM can be understood as a continuous, simplified representation, through external surrogates,oftheinternal temporalstructureofthemachine.
Because the A input and O input values are correlated with each other, their content is in effect erased. Because each memory slot in A/O LTM encodes so little information, it becomes in effect a series o f marks which are themselves marked by the change in vocal timing. Sensory input is organized within a structure generated in relation to an internal state change. The A/O LTM models both external and internal time within asinglestructure. Evenmoresignificantlythelevelofabstractionoftheinformationused by the A/O Evaluator has allowed the creation of a model of the structure of time divorced from its content (defined in relation to a unit o f difference, not an instant in time). Our problem now is to figure out how to use this structure.
14. 13 Negative entropy
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
605
? The A/O LTM gives our machine the ability to make a long range prediction. We must assume that at least one machine can outlive a number of its matches. If so, the only pattern a memory analyzer o f A/O LTM could identify would be that every vocal synchronization change will be followed by another change after a period o f time. Because the content of the A/O LTM is so poor, if a change in vocal synchrony occurs and recalls the memory of past changes (that is, the pattern of such changes), it will recall the temporal structure embodied in the LTM. Thus, whenever a new match occurs a long-range prediction would be possible: this vocal change will be followed by a series of differences (temporal units) which will themselves end in a new vocal timing change.
The structure supporting this prediction (the moments represented by the O Difference
Detector) will be in place until this expectation (the next vocal timing change) has been met, at which point it will begin again. Attached to the A/O Evaluator will be a memory recall agent that will work in parallel with the Evaluator, but will also organize both A and O input within the temporal structure imported from the A/O LTM. We have a future.
14. 14 Re-building ourselves in the other
This leaves us with two problems however. How can we integrate the other
perceptual, evaluatory, and memory agents into this structure? And what is the significance o f the two states (match and unmatched) within the temporal series in the A/O LTM?
The notion of the future, dependent as it is on a society of machines (identical in everything but life-span), requires a society of different sensory agents. Is it necessary,
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
606
? however, to integrate the O LTM, with its patterns and short-term predictions, within the long-term future of A/O agent? If our machine receives optical information that triggers a short-term prediction, we might want this to signal the A/O Evaluator. But the A/O agent could not understand this kind of pattern (between numbers of objects). Such a pattern would not be reflected in its degree of change information. The only patterns it could recognize, other than the change in vocal synchronization, is the shift from its shunting of matched to unmatched information to LTM and a correlation between a matched auditory inputandeveryfivesignalsfromtheOpticalDifferenceDetector. IftheAagentcould activate a pattern recognition agent to evaluate the unmatched A STM that would be sensitive to more abstract patterns o f difference (every degree o f change 2 is followed by a degree o f change 4), and if we pretend that these patterns are present in some districts, then our two memory structures could be integrated easily by running them both through
the same memory recall agent. The existence o f such patterns in the machine world would indeed be convenient, but then so is the bipolar structure of water.
We have three sub-agents, constituting three internal time scales, not fully exploited in modeling time: the (1)V Difference Detector indicating a change in vocal timing, the (2) A/O Evaluator's awareness that every match happens after 5 optical moments, and the (3) A/O Evaluator's (one o f its internal agents) ability to detect the shift from matched to unmatched outputs to memory. (1) and (3) constitute long-term regularities, while (2) is a short-term prediction that can act as the structural content organizing the temporal structure enabling long-term predictions.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
607
? The machine's change in (1) vocal timing finds its analogue in the death o f the external matched signal: the first registers as an internal change; the second as an external change. We still do not know if our machine understands this as the death o f its machine mate. An existential understanding ofthe loss ofthe matching voice would require (ifwe ignore the lack o f a mammalian brain, or some might say any brain) an analogic equivalence between the A/O Evaluator shift to unmatched memory output and the death o f itself. It seems unlikely that a machine without a more developed symbolic life could
understand this loss. The predictive power of this kind of system, however, remains stimulus dependent. When the stimulus vanishes, the future disappears.
14. 15 Thepursuitofdeath
If we could make a higher level analyzing agent to determine and chart the larger patterns of change in A/O LTM [(3): matched to unmatched], we might be able to stimulate the machine into a weak awareness of the subjectivity of another machine encoded in the A/O LTMs meta-temporal identity. The A/O Evaluator sends a signal to
this agent when the shift from matched to unmatched occurs, and this signal causes this Pattern Recognizer to recall previous shifts and their defining vocal timing domains. The Pattern Recognizer could construct an information structure that charts the internal state changes against the representation o f catastrophic loss o f the external matching voice.
The information such a chart could encode is rather simple, and the prediction that follows even simpler: every vocal timing change will be followed not only by another, but also by the loss ofthe not-I representation. Because the representation ofthe not-I is used to
with permission of the copyright owner. Further reproduction prohibited without permission.
? determine the actions o f the machine (it stays and observers), its loss results in a change in action (the machine now moves). Before moving can be a form o f mourning, it must also representalossofbeingforthemachine. Ifweremovethemachine's eyesonthesideof its head, it will lose its primary temporal portal. It will still have its auditory input (and STM) by which it can chart a different temporal order, but the loss of this primary information will be recorded with the shift from matched to unmatched in the A/O LTM and will be noted by the A/O LTM Evaluator. This sensory loss, and the time/ memory scale it constructs, can stand for, can symbolize, the loss o f the external voice and the change in vocal timing. The prediction that a loss o f the external voice will necessarily follow a change in the machine's own vocal timing will also be tied to another internal
effect: thecommandtomoveanditssubsequenttemporarylossofsightandthetime- scale that sight generates. This causal structure (Cause [external loss] and Effect [internal loss]) mirrors the vocal timing change and is nested within the basic structure generated from that timing change. Our sense o f the future is growing more complex.
14. 16 Learning from the future
The bifurcation o f the memory and predicative domain defined and signaled by the
vocal timing change sets up a meta-time scale at which level internal symbolic exchanges begin to make some kind o f sense. The O agent wants to be able to predict when it will fail, and when it's short-term predictions will be irrelevant, while the A/O wants to be able to predict when it will have to shift its memory output from matched to unmatched. Our machine could not discover such detailed information, nor could these agents as they now
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
609
? stand actually want anything. They would first have to need this information as a means to act. But what is important here is the conceptual problem this presents. For the Optical Agent the loss of the external voice could stand for its loss o f sight, and because it happens prior to its effect, it could serve as a short-term prediction (generated out of a long-term prediction structure). This represents an internal symbolic communication, where one loss can stand for another. The signal itselfwould go from the A/O Evaluator to the O Agent. This signal is not, however, caused by an algorithm: if shift to unmatched output, send a signal to O that loss of sight is imminent. The meaning ofthe shifttounmatchedoutputisdeterminedbytheA/OLTMPattern. Itconcludesfrom previous patterns that the shift means a loss o f sight. This is a conclusion based on an internal analysis (abstraction) from a lower level abstraction (A/O LTM) which is in turn
generated from lower level sensory modeling. This is not the most elegant of control structures (but neither is much of our learning compared to "lower" animals: human weavingcomparedwithspiderweaving). Ithastheadvantage,however,ofmakingall predictions a function ofmemory (they are learned), and not of pre-programming.
14. 17 Inside a Chinese box: when structure becomes content within larger structures
A matched memory state (any input of value 0) within the A/O LTM, re- interpreted in terms o f the time scale generated from the Optical Agent, stands for 5 Optical moments. If a match has been found, then every five O moments will result in a matched input to A/O LTM. The Pattern Recognizer attached to A/O Evaluator
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
610
? functions as a symbolic catalyst linking the two time scales. The Pattern Recognizer has the power to include in any memory state within A/O LTM a representation of the 5 O moments. (Unfortunately,ourmachineistoosimpletoneedthiskindofsymbolic transfer. Without this need and use all one can say is this machine has the conceptual structure that could allow this integration o f two time-scales as a form o f its temporal consciousness).
14. 18 The I as the not-I: using the future to build other machines
Things are not so easy for the A/O LTM Pattern Recognizer (not to be confused
with the A/O Evaluator's Pattern Recognizer). It can learn to expect the loss of the matched input, but not when, except to the degree that it can know it will happen before the next vocal timing change and before the next loss of sight. In fact the initializing vocal change and the loss of sight define the domain of the matched memory. At the level of abstraction found in the A/O LTM Pattern Recognizer this is enough information to constructourmeta-temporalidentityinexistentialterms: adomaindefinedbytwo internal state changes and consisting of a model of internal perceptual states, but which are
notthoseperceptualstates. Thelossofsightcanstandforthelossofthisdomain. A feedback loop from the A/O LTM Pattern Recognizer could signal the loss of the matched domain to the A/O Evaluator, and thus eliminate the previous hard-wire instruction to shift its output to LTM when it no longer computes a match between A inputandOinput. ThetranslationofauditoryinformationintoalanguagetheOagent can understand ("you will lose sight"), and the reverse translation ("you must shift to
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
611
? unmatched memory"), in spite ofthe physical improbabilities, constructs the symbolic structure that allows for vocal timing changes to stand for a future state o f the machine. Because this state is in the future, it is a not-I. Because this state, while recognized as the representation of the machine's internal state, is built from and also represents an external input, it is a not-I. It is easy for the machine to recognize itself in the auditory external input (the not-I as I). But much more difficult to recognize the I as a not-I. The future provides the "distance", the abstraction that allows the I (the internal state of the machine) to function as a not-I equivalent to the not-I it already recognizes as a representation of itself. Thus, the machine cannot quite recognize the death of the other machine, but it can recognize the loss o f the external auditory signal that it uses to model its own internal state with the new and different life (internal state) the future offers it. It can believe in the possibility of the future.
This is kind of self-deception, used like fiction, to jump into the possibilities that logic cannot capture. It makes the machine what Stanislaw Lem might call a self- inductingmachine. WhenErgtheSelf-Inductorwounduptheclockprincesswiththekey to her life, stolen by a terrible human paleface, he had to play with time (and make up stories) in order to hide the fact that he had simply fabricated a new key instead of pursuing the mad human through space.
Our machine builds (or represents) its world, including other machines, within a hierarchy of abstractions, in order to establish the analogic connections that will allow it to identify (self-induct) itself with an other by self- inducting itself into the future not as itself, but as the not-I represented in A/O LTM.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
? Our machine builds a temporal map in order to make long term predictions. The Optical memory recall Agent can be nested within the A/O LTM recall Agent, thus integrating both time-scales within a single time sense. This is possible because the temporal structure that the A/O LTM Recall Agent remembers is both continually active anddevoidofanycontent. Eachmomentissimplyastructuralmark,and,becauseitis defined by degree o f change values, it recognizes not single phenomenal states, but the dynamic two state moments stored in O STM and O LTM.
In a more complex organism like human beings, however, this future structure, built out of internal changes caused by external inputs that simultaneously represent an other and the self (that is, there is a distinction between other and self, but the other can sometimes be used to represent the self), can abstract others to the point that they can stand for oneselfnot only in the future possibilities of life, but in the future possibilities of death. Outofthisallculture,includingtechnologyandlanguageisbuilt.
1This exercise can seem as a conceptual counterpoint to Heidegger's analytic description of our being towardthefutureinBeingandTime. Iamlookingatdifferentwaysinwhichbeing(orbecoming)canbe described. Iamnottryingtodescribebecomingatallhere. Rathermymachinesaremeanttofunctionas such a description.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
613
? Epilogue: But what have I denied the existence of?
I claimed earlier that my machine(s) would serve as an allegory about how ontological limits describe ways ofmaking sense through the process ofbuilding. In what way are these machine(s) allegories? Commonly, an allegory is an ordered set o f relations withwhichorintowhichatextorsituationiscorrelatedortranslated. Inthissense allegories are applied. Such applications are fraught with dangers and confusions. A possible interpretation is not always probable. Furthermore the application o f an allegory can confuse the interpretive relations within the allegory with a causal explanation of events or histories (e. g. Freud or many political allegories). My machine allegories are very different. In fact in themselves they are not meant as allegories at all. The process of making them, o f correlating aspects o f my experience o f the world in relation to a clear set of ontological limits should be understood as allegorical. I am not writing an allegory of things or o f the world these things inhabit but an allegory o f the making o f these things and their world. The allegory does not have content, but the process of making the machine(s)does. Thisisanallegoricalprocessinwhichtheengineeringlimitsconfronted in building the machines opens the world as a system o f limits, organized around
causation, but directed at removing black boxes (homunculus or God or soul or the unconscious). Suchremovalcreatesanallegoryoflimitswithinwhichwebuildaspectsof sense as a way o f orienting ourselves within the world and or experience.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
? My machines have few places to go, confined as they are to their world, but they haveawhentogointoorremainwithin. Thiswhenisasyntaxwithaverysimple semantics. Their processing builds a temporal envelope whose surface is constructed through the mutual correlation of a machine's own actions (vocalization) with its recognition o f another and with the limits o f the world (not to mention the perversity o f its creator). In many ways the construction o f a similar envelope has been the project o f this dissertation. I have attempted to construct a context within a set of texts in order to make visible the ways in which our negotiations within and towards sense and nonsense generate temporallimits. Howweinhabit'I'and'our'expressesourinvolvementinlanguageat the level o f symbolic functioning that confuses the future and the present (time) and meaningandtheword(grammar). Wefunctionasandwithinfragmentsoftimeand language through which we expose and disguise the limits of how we are, not just what or
who we are.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
615
? Bibliography
Abrams, M. H. The Mirror and the Lamp: Romantic Theory and the Critical Tradition. NY: Norton, 1953.
Adams, Henry. The Education o fHenry Adams. Ed. Ernest Samuels. Boston: Houghton Mifflin Co. , 1918, 1946, 1973.
Adorno, Theodor W. Negative Dialectics. Trans. E. B. Ashton. New York: Continuum P, 1966.
Adorno,Theodor. Quasiunafantasia: EssaysonModemMusic. Trans. Rodney Livingston. NY: Verso, 1992.
Ammons, A. R. Shere: The Form o fMotion. NY: WW Norton, 1974.
Arbman, Ernst. "Untersuchungen zur primitiven Seelenvorstellung mit besonderer
Rucksicht auflndien. " In Le Monde Oriental 20 (1926) 85-222 and 21 (1927) 1- 185.
Aristotle. Aristotle On the Soul. Trans. W. S. Hett. London: Heinmann. Loeb Clasical Library, 1936.
_______. Aristotle's Metaphysics, Books Zeta, Eta, Theta, Iota. Trans. M. Furth. Indianapolis: Hackett, 1982.
_______. The Physics. Trans. R. P. Hardie and R. K. Gaye. In The Basic Works o f Aristotle, ed. Richard McKeon (New York: Random, 1941).
Ashberry, John. Flow Chart. NY: Alfred A. Knopf, 1992.
____________. ReprortedSightings: Art Chronicles 1957-1987. Cambridge: Harvard
UP, 1991.
Atherton, James S. The Books at the Wake: A Study o fLiterary Allusions in James
Joyce's Finnegans Wake, Carbondale: Southern Illinois UP, 1974.
Aquinas, Thomas. Summa Theologia. Latin and trans. Cornelius Ernst O. P. (NY:,
1972); Summa Theologiae: Cura Fratrum eiusdem Ordinis, 5 Vols. (Madrid: Biblioteca de Autores Cristianos, 1994).
Augustine. Confessions. Trans. Henry Chadwick. NY: Oxford UP, 1991.
________. De Trinitate. In Augustine: Later Writings. Ed. John Burnaby. Philadelphia:
The Westminster P, 1955.
Austen, J. L. "A Plea for Excuses". In Philosophical Papers. Oxford: Oxford UP, 1979. Bach, Kent. Thought and Reference. Oxford: Clarendon P, 1987.
Barnes, Jonathan. The Presocratic Philosphers. Boston: Routledge and Kegan Paul, 1979. Barth. Karl. The Epistle to the Romans. Trans. Edwyn C. Hoskyns. Oxford: Oxford UP,
1933.
_________. The Theology o fJohn Calvin. Trans. Geoffrey W. Bromiley. Grand Rapids:
William B. Eerdmans Pub. , 1995.
Beckett, S. et al. OurExagmination RoundHistFactificationfor Incamination of Work
in Progress: James Joyce/Finnegans Wake: A Symposium. NY: New
Directions, 1929.
Benjamin, Wlater. The Origin o f German Tragic Drama. Trans. John Osborne. NY:
Verso, 1977.
______________. "Re the Theory ofKnowledge, Theory ofProgress," In Benjamin:
Philosophy, Aesthetics, History. Ed. Gary Smith. Chicago: U of Chicago P, 1989.
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.
616
? Berger,Harry. SecondWorldandGreenWorld:StudiesinRenaissanceFiction-Making. Berkeley: U ofCaliforniaP, 1988.
Benveniste, Emile. Indo-European Language and Society. Trans. Elizabeth Palmer. Coral Gables, Fla. : U ofMiami P, 1973.
______________ . Problems in General Linguistics. Trans. Mary Elizabeth Meek. Coral Gables: U ofMiami P, 1971.
Bishop, John. Joyce'sBookoftheDark. Madison: UofWisconisnP, 1986.
Bishop, Elizabeth. The Complete Poems: 1927-1979. NY: Farrar Straus Giroux, 1979. Blake, William. The Poetry and Prose o f William Blake. Ed. David V. Erdman. Garden
City: Doubleday, 1965.
Blanchot, Maurice. " The Athaneum" In The The Infinite Conversation. Trans. Susan
Hanson. Minneapolis: U o f Minnesota P, 1993.
