Information theory at the keyboard — Henry Quastler’s “Studies of Human Channel Capacity” (1956)

Eamonn Bell

22 January 2019 @ MPI-EMEA


When Claude Shannon’s “A Mathematical Theory of Communication” was published in 1948, engineers had already made ample use of the new quantity of information defined therein to build more secure and more efficient communications systems. A critical–interpretative tradition that sought to expand the explanatory domain of information theory began almost immediately. In a short but profoundly influential essay, Warren Weaver asserted the broader applicability of Shannon’s theory to a variety of fields of cultural production, including “music, the pictorial arts, the theatre, [and] the ballet.”1 Revisiting reports of laboratory research into musical behavior dating from the first decade of the age of information, I explore whether such optimism about the generality of Shannon’s theory was justified.

In this talk, I focus on the physician Henry Quastler’s “Studies of Human Channel Capacity” (1956). In one series of tests, carried out at the Control Systems Laboratory at the University of Illinois Urbana–Champaign, Quastler asked pianists to sight-read scores that were generated according to the outcome of controlled random processes.2 Both the set from which pitches were drawn and their rate of presentation were varied independently, “coaxing the subjects into greater and greater speed until they were obviously way beyond their capabilities.”3 Quastler quantified his subjects’ performance as “information transmission rate”, measured in bits per second, thus proudly using Shannon’s definition of information to precisely describe the limits of his subjects’ abilities across a variety of tasks, including sight-reading at the piano.

I argue that, despite (or perhaps, because of) its many methodological oversights and flawed assumptions, the turn to information reflected in Quastler’s study views pianism a specific manifestation of a more general capability for information “transmission” at or close to a putative human “channel capacity”. This assumption motivates Quastler’s description of a hypothetical modular framework for understanding “simple sequential tasks”, on which the influence of cybernetics is keenly felt. In designing this research program for the study of musical behavior under the sign of Shannon’s information, Quastler’s research evidences the dual of what Brian Lennon has recently called Weaver’s “culturalization of information”: the twentieth-century’s informatization of (musical) culture.4



In 2013, two researchers at the Max Planck Institute for Informatics, Anna Feit and Antti Oulasvirta, presented a remarkable demonstration at the annual ACM Conference on Human Factors in Computing Systems: a new kind of musical typewriter called PianoText. Feit, then a doctoral student, claimed that a trained pianist “can produce 12 notes/sec easily.”5 Thanks to algorithms that discover an efficient correspondence between the keys of the piano and the English alphabet, such a user—once familiar with this mapping—would be able to achieve corresponding typing rate of 144 words per minute.


This technical system is one of a kind that motivates my own doctoral research, because it sits at the intersection between music and information technology. Feit’s work is especially interesting because it claims to leverage in expertise one domain (music) in order to make another more efficient (typewriting). In contradistinction to the countless computer systems that aim to open up the expert domain of music to the otherwise musically untrained average computer user, the PianoText project explores what it is like to take advantage of the experiences of trained pianists. In this case, music is source of inspiration for new human-computer interaction paradigms. More generally, their research exemplifies the flow of ideas between music and the other disciplines, outlining the stakes for taking musical performance seriously as a repository of uniquely skilled behaviors.

A full account of the apparatus that makes this kind of technology transfer possible requires an approach that is interdisciplinary. After all, the keen musicologist sees in such a demonstration the long history of keyboard pedagogy concerned with the perfectibility of performing bodies. On the other hand, the historian of communications technology sees something different: the lineage of chorded alphabetic keyboards that stretches back to the first stenographic typewriters in the middle of the nineteenth century. Since PianoText also depends on the measurement of both letter frequencies in the English language, as well as computing the note-transition probabilities in a representative corpus of tonal music, it also draws on the last 75 or so years of research into natural language processing and, even, the relatively new field of Music Information Retrieval.

In this paper, I suggest that speculative computing applications like PianoText are only thinkable when music and typewriting are both conceived of as a kind of information processing. This idea is many decades older than this demo. This particular attitude toward musicianship is the manifestation of an informaticized view of musical production that reaches back to the earliest applications of information theory, as transplanted outside of its original discipline of communications engineering.

The idea that underwrites PianoText—that musical behavior and typewriting are simply special cases of more general human capacity for information processing—comes out of the messy profusion of applications of Claude Shannon and Norbert Wiener’s transformative theory of information, which began to be disseminated among engineers after the end of the Second World War.

I point to research conducted in the first half of the 1950s at the University of Illinois by Henry Quastler, which demonstrates this notion in action in the behavioral psychology laboratory. Reporting on the results of his experiments, to which I will shortly describe in more detail, Quastler makes the following remarkable observation about pianism, using the terms of information theory:

Piano playing is a sequential task which involves high rates of information transmission. A piano virtuoso generates information at a rate which is very impressive indeed; it is possible that that the performance of a piano player may approach the peak of human capability.6

Over the next twenty-five minutes or so, I will discuss the disciplinary and experimental context for this quote, spend most of the time discussing Quastler’s work at the University of Illinois. Before I do so, I’ll quickly outline the tools he used.

Early days

Claude Shannon’s work on information theory, which he carried out during a stint at Bell Laboratories in the latter years of the 1940s, built on earlier definitions used to quantify information transmission in telegraphy. Shannon proposed a generalization of an earlier mathematical definition of information so that it was, as William Aspray has put it, applicable to “any system, physical or biological, in which information is being transferred or manipulated through time and space.”7 This measure of information is known variously as Shannon information or Shannon-Wiener entropy. Shannon pointed out that many problems in communications shared some common features: a transmitter, which sends messages in the form of signals over a channel that may or may not be subject to distortion, and a suitably equipped receiver apparatus which detects differences in the signals, resolving them into the original messages. This is what I understand “information theory” to mean for the purposes of this talk: a preference for the description of processes in terms of this schematic overview (a model), and the use of Shannon information in empirical or theoretical studies of these processes (a measure). The model and the measure, taken together, allowed research to describe communications processes so that the amount of information thus transmitted was quantifiable, drawing on the statistics of stochastic processes to define information content as a mathematical quantity, now conventionally measured in “bits”.

The generality of Shannon’s definition of information provides one reason that a critical–interpretative tradition could be initiated almost immediately: in the landmark re-publication of this paper one year later as A Mathematical Theory of Communication (1949), Shannon’s essay was preceded a short but profoundly influential essay by Weaver.8 Weaver speculated about the broader applicability of information theory to a variety of fields of cultural production.9:

The word communication will be used here in a very broad sense to include all of the procedures by which one mind may affect another. This, of course, involves not only written and oral speech, but also music, the pictorial arts, the theatre, the ballet, and in fact all human behavior. […] The language of this memorandum will often appear to refer to the special, but still very broad and important, field of the communication of speech; but practically everything said applies equally well to music of any sort, and to still or moving pictures, as in television.10

Brian Lennon, writing just last year (2018) about the rise of quantitative style analysis and statistical authorship attribution studies in the literary humanities, referred to this process as Warren Weaver’s “culturalization of information”. By putting into operation a research program for the study of the music under the sign of information theory in the decade that followed the field’s landmark publications, psychologists and composers would evidence the dual of this process: the informatization of (musical) culture.

The discovery and use of information, it must be pointed out, is not exclusively tied to the growth of computing. In other words, information-theoretic conceptions of music need not be specified with respect to a particular computer system or artificial environment. As William Aspray has argued, the history of information should be decoupled from the history of digital computing (sometimes confusingly called “informatics”). The scientific conceptualization of information resulted from the consolidation of decades of research in electrical engineering, mathematical logic, physics, psychology, and neurobiology, all of which predated the design and manufacture of digital computers as we would recognize them today.11 Norbert Wiener, the MIT mathematician whose writings were extremely influential on Shannon, would advocate for the term “cybernetics” to refer to the opportunistic borrowing from these fields to describe complex organic and inorganic systems. As such, Henry Quastler’s research, to which I now turn, represents one of the earliest attempts to apply cybernetic techniques to the study of musical behavior, many years before cybernetic principles filtered into the production of interactive and reactive art in the early 1960s.

Human information processing in Urbana

One of the earliest uses of information theory to characterize explicitly musical behavior can be Henry Quastler’s studies of human “channel capacity” at the Control Systems Laboratory at the University of Illinois, Urbana–Champaign.12 Quastler was an Austrian physician with a specialism in radiology, who emigrated to New York in 1939. In 1941, Quastler moved to the cornfields of Urbana where he practiced medicine until his appointment to a assistant professorship in physiology at the Control Systems Laboratory, in 1949.13

Though his name is probably unknown to the majority of music psychologists today, Henry Quastler was by no means a marginal figure in his field. The historian of science Lily Kay has credited Quastler with the introduction of concepts from information theory into molecular biology that have now become naturalized as commonplaces of the field. She traces first use of cryptographic and scriptive metaphors in molecular biology back to Quastler’s early research into genetics. Although Quastler’s insight was incomplete, and ultimately made obsolete by later research, Kay argues that Quastler’s early work on the genetic code nevertheless led to the widespread of communication-related metaphors by molecular biologists: words like “decoding”, “transcription”, “instructions” persist today, supporting the powerful metaphor of “information as the ontological unit of life”14

At University of Illinois, Quastler headed up the Biological Systems Group, one of the founding research units at the Control Systems Laboratory. By the time he was appointed to the Biological Systems Group, Quastler was very familiar with the work of Shannon and Wiener on communications. He saw fit to adjust his focus away from information-theoretical applications in molecular biology and instead on to human behavior considered at the macro-scale: the information-processing behaviors of individuals and organizations. A report from 1955 summarizes the work of the Biological Systems Group into four main strands:

  1. bio-systems
  2. man-machine systems
  3. human information processing
  4. applied mathematics

In this paper, I focus on work that was carried out under the human information-processing strand; this research strand, Quastler writes, made up the “main body” of the work of the group from 1951 to 1956.15 Quastler’s group was especially interested in determining the limits on the capabilities of human beings in a variety stressful situations. This topic was of general interest to the growing community of psychologists who saw a useful quantitative tool in information as well as to military planners and industrial researchers, who were interested in computing the maximum capacities of technological systems that depended on human beings. As research in the “man-machine systems” strand had shown, “not nearly enough was known about the properties of the basic components” of teams of human beings—that is, “a single man processing information”.16

Quastler’s experimental design was undergirded by the apparent generality of information; he assumed that

human information processing can be represented by a limited number of models, independent of the particular kind of information processes, then one can attempt to determine the human capabilities in a general way, not using those tasks which are ultimately of interest, but those that are most accessible.17

Quastler was less interested in quantifying information transmission rates in these particular tasks than in characterizing a larger slate of behaviors—“perceiving, filtering, remembering, correlating, learning, selecting a responses”—that were subject to an information bottleneck at the input and output steps.18 The performance of humans on specific, easy to measure, and operationally more convenient tasks could be used to infer limits on their behavior more generally.

Quastler summarized his group’s recommendations as follows:

If faced with a particular situation with (approximately) known informational requirements, and the problem of determining whether or not it will over-tax a man’s capabilities, search for a similar situation among well-learned activities such as those treated in this report.19

Quastler set out to study the performance of individual human subjects that were idealized, in a way, as black boxes: subjects would be presented with input “information”, int the form of some visual or auditory stimulus, sometimes called “displays”, and were asked to respond with output “information”: usually the value indicated by the display. With these assumptions in mind, Quastler developed a battery of activities that were designed to shed light on limits of human ability in terms of information transmission. He chose behavioral tasks that appeared to conform to three criteria.20

First, all inputs should come from a single source: the task must be possible with reference to a single kind of stimulus, be it a dial, characters on a sheet to be transcribed, or in musical notation. Second, all output choices are “mechanical”, which is to say that no reflection or consideration is required to choose the correct output response, and that there is only one correct output for each input. Finally, all tasks should be “thoroughly familiar” to the subjects; elsewhere, Quastler describes his focus on “overlearned” tasks, which connotes a skill in the task bordering on automatism.21 Quastler argues that these three constraints ensure that the information processing demands of the tasks are kept to a minimum, so that the subjects’ performance statistics can be interpreted as relating to theoretical maximum information transmission rate—in other words, throughput or channel capacity—and not a bottleneck elsewhere in the chain of information flow.

A battery of tasks

This table summarizes the tests that Quastler carried out, the results of which are described in a series of unclassified reports, recently made available at the University of Illinois.22 The activities are broken down into three groups: sequential tasks, “flash recognition” tasks, and scale-reading tasks. The first group of columns shows the independent variables that were explored using in each task farming. Notably, the “piano playing” task had the most scope for parametrization of all the tasks. As we will shortly see, the affordances of musical notation and skilled musical listeners allowed the experimenters to independently vary the speed, order of complexity, and range of motion used in stimuli.

Not only did the tasks have to be on a par with real-world activities like reading radar displays or responding to air-traffic control commands, they also had to be comparable to each other. Quastler seemed more certain that his experimental results reflected some empirical reality when his putative channel capacity calculations agreed with each other across a variety of modalities. This motivated Quastler’s preference for Shannon-Wiener information over other, less sophisticated statistical summaries (such as mean or median accuracy scores) when computing the performance of his test subjects in each of the tested activities. Information theory provided a common unit—the “bit”—which could be used to quantify the variety of ideal behaviors.

The appearance of this unit, of course, prefigures its use today to measure data in storage and in transit in actual computer systems. Here, however, it measures the amount of information accurately transmitted in terms of the number of hypothetical yes/no decisions required to definitively distinguish between the possible values of the message, independently of machine or apparatus. Shannon’s information allowed Quastler to summarize the performance of humans in his experiments—as superficially diverse and distinctive as the various exercises seem to us now—as follows:

This approach furnished Quastler’s group with a method for assessing the limits on human performance in real-world situations without the need for expensive and time-consuming simulations or for potentially risky “field tests”.24 Thus, innocuous tasks like typewriting, dial reading, and the recognition of flashcards were deployed in the lab as ways to measure proxies for other measures of human behavior. The validity of this research, Quastler admitted, hinged on assuming that these specialized behaviors were in some way commensurate with the other activities that his military stakeholders were interested in. So too with PianoText, as Feit emphasizes that the design of her algorithm promotes the “transfer” of skills from the musical domain to the non-musical. On this point, both Quastler’s and Feit’s researches trade in a similar perspective on piano performance. It is a vision of pianism that is departicularized, studied from a point of view that looks through the contingencies of musical expertise: a flattened view that considers musical performance as just another skilled activity—like typing.

There is a caveat, of course: not all of these upper limits can be interpreted as channel capacities. “There definitely is”, Quastler writes, “more than one kind of limiting factor”, and this is specific to the kind of task in question.25 For sequential activities, such as typing and piano sight-reading, these limits are modulated by the target speed of the activity, the range of symbols over which the subject must range, or their product (whichever is lowest).26 In order to determine the maximum information transmission rates in the piano-playing task, Quastler asked three trained pianists to sight-read scores that generated were generated by hand according to the outcome of controlled random processes.27 Some examples of the kinds of scores that were used in these tasks were shown in Figure X, along with the nominal information content of each individual key stroke.

Low information-per-key statistics correspond, following Shannon’s definition of information, to scores generated by random selections from smaller selections of possible notes. In all examples, the notes are selected according to uniform distribution over a set of possible keys of the piano: in each stimulus type, each note has equal probability of occurring, with no conditional dependencies between successive notes. What Quastler would argue to be the information content of an individual note in the stimulus is therefore proportional to the size of the total set of possible pitches in each stimulus. The notes played were intended to be strictly isochronous and of equal duration, but the example passages show some a mismatch between the notation and the intended stimulus.

During a session, the set from which the pitches were drawn and the rate of presentation were varied independently of each other, “coaxing the subjects into greater and greater speed until they were obviously way beyond their capabilities.” [Quastler,28 p. 14; discussed in Fred Attneave,29 p. 79–80. Quastler derived his maximum information transmission rates from both the speed and complexity of the stimuli, but also from the pitch accuracy of the sight-reading, using tape recorders and the three pianists to determine each other’s accuracy by ear. Quastler measured the pianists’ performance in bits per second, using Shannon’s quantity of information to characterize the maximum limits of his subjects’ ability to sight-read.

Quastler justifies the use of randomly-generated sequences of pitches in his experiment by describing how the musician learns both specific pieces and a sense of musical style in terms of information acquisition. A “virtuoso” playing a piece that they already know has a high information output rate, due to their familiarity with the piece, but this does not “depend on information acquired at the time of playing”.30 A piece that is “unknown but of a familiar kind”, poses a related but distinct problem: experienced musicians will draw on their knowledge of similar pieces, and “use old information as well as new, and guess some notes without having read them”. The subject’s prior familiarity with musical style—in other words, “old information”—would have to be controlled for in Quastler’s experiments. Quastler suggests that “by studying the statistical structure of the musical style, one could establish how much old information the subject could use, at best, to facilitate the acquisition of new information”.31

However, Quastler concedes that determining how much old information the subject actually uses is difficult to evaluate. Quastler concludes this confound means that an experiment making use of an already-learned piece or a piece in a familiar style cannot be used to measure the pianists information processing capacity. Subjects would draw on their reserve of “old information”, accrued through memorization or familiarity with a musical style, separately to the information supposedly deposited in the “input” stimulus, and the transmission rate would thus reflect not only the information that was to be found in the score but also the information that corresponds to the subject’s prior knowledge of the style. Quastler thus proposes the following solution:

One can get around this problem by setting up experimental situations in which all the information in the text is new. This certain to be the case when the script used is a random sequence of notes. Such a script is easily made up with the help of a table of random numbers. In this situation there is no doubt about the information input.32

The scores shown here, that were generated according to the outcome of a random process, are supposed to control for the subject’s knowledge of the repertoire (in the musical sense).

Discussing the results of these experiments, Quastler noted that the limitations on the rate of information processing are not caused by limitations on the rate at which sense data is provided to “peripheral input mechanisms”.33 For instance, the optic nerve was hypothesized to have a channel capacity of several orders of magnitude greater than that implied the maximum transmission rates observed in contemporary studies. The apparent richness of visual input, when viewed as data, suggests that humans can ingest information much more quickly than it appears to be able to process it. Quastler rather naively argues that limitations on the output rate—the act of striking the keys in the piano case—are not due to “mechanical difficulties”, since he notes that with rehearsal, the output rate could be improved.34 This leaves Quastler to conclude that “the mechanisms which limit the observed performance must be connected with the speed of processing information”.35

Although he was not primarily focused on accounting for the mechanisms behind these processes that were involved in the various transmission tasks, Quastler did discuss them briefly, decomposing the actions of his piano-playing subjects into the following “components”:

  1. visual input
  2. symbol recognition and selection of action
  3. coordinated planning of motor action
  4. (actual) motor action
  5. error detection36

Figure Y shows Quastler’s schematic representation of a hypothetical cognitive model for the sight-reading task, showing how the symbols of a musical score are detected and processed by the five “component” systems listed above. Quastler explains:

The physical events (scripts and sounds) are projected on bands; the boxes represent mental processes, or rather, the abstractions here used to describe areas of mental activities; each box represents a complex system of action37

The interconnections shown between the various cells represent pathways through which information might be said to flow; the question of precisely how this “information” might be coded for as trains of impulses or as other neural representation is deferred. This was never Quastler’s objective.38 But this figure is crucial to my claim about what Quastler’s project represents, not because of the particular model of piano sight-reading is novel or accurate, but because of its claims to generality. Quastler writes that “the same type of diagram could be used for many other serial activities—such as walking, reading, or speaking”.39 This is the crux of Quastler’s work: a convergence between the approximate estimates of human channel capacity across the wide range of activities—typewriting and piano sight-reading included, as we have seen so clearly in Feit’s PianoText—pointed to the possibility of a general framework for first quantifying, and then describing, these “overlearned” activities in terms of a new explanatory paradigm: the human as an information processor.

Conclusion: Quastler and computer music?

To conclude, I reflect on where I think Quastler’s work should sit in the history of music computing, as I have suggested it should. Given that the randomly-generated scores are almost trivial in design and are used entirely instrumentally as experimental stimuli, should an account of Quastler’s research therefore be excluded from the history of music because it so obviously lacks “real” music, or a “real” composer? A similar question is relevant to Feit’s PianoText: what place do such idiosyncratic—yet highly technologically provocative—uses of musical instruments and expertise have in the account of mainstream musical behaviors?

The answer, in part, depends on what kinds of things we are willing to admit as the agents of music history. A contemporary materialist turn in musicology aims to supplement the traditional subjects of music history—musical works, the inner lives of composers, and their performers—with a focus on material artifacts, conventionally considered to be inanimate for the purposes of history. As Emily Dolan has eloquently put it, such a reorientation serves

the purposes of demystification—turning our attention away from the enchanting musical works blossoming in the Adlerian garden and toward all the things that have made them possible: the invisible laborers, their tools, their techniques.40

In that spirit, I want to recall that from 1952 onward, researchers at the Control Systems Laboratory had access to ILLIAC: a vacuum-tube computer which was built according to the designs of the influential stored-program architecture, first developed by John Von Neumann at Princeton’s Institute for Advanced Studies but never used. ILLIAC was the first computer built and entirely owned by a US educational institution.41

In a deal negotiated with the US Army’s Ballistic Research Laboratory, the University of Illinois agreed to design and build a stored-program computer based on Von Neumann’s unused designs, buying two of each part required by the build. The result, rather unusual for the time, was that two identical (and therefore completely compatible) digital computers were completed in Bloomington in 1952: ORDVAC, which was disassembled and sent to Maryland to be used to compute missile trajectories; and ILLIAC, which remained at the University of Illinois and operated for a decade there before being replaced by another custom-built machine in 1962.42

ILLIAC plays an important role in the history of music informatics: it was this computer that chemist–composers Lejaren Hiller and Leonard Isaacson used to compose the Illiac Suite during the period between September 1955 and 1956, perhaps the first significant composition entirely generated by a digital computer.43 The Illiac Suite, a four-movement piece for string quartet, showed how so-called Monte Carlo processes could be used to impose stylistic constraints on randomly-generated music so that it at least resembled known forms of composition. Hiller would go on to later collaborate with John Cage in the production of the multimedia work HPSCHD (1969), which used the successor to ILLIAC to generate audio and visual components for an immersive night of entertainment with dozens of performers and hundreds of playback and projection elements.

Much like Hiller and Isaacson had done, Quastler used the output of controlled random processes to produce musical scores. As we know, of course, these scores were not viewed as self-standing musical compositions. Rather, they were to be used as visual stimuli in the study of human information-processing capacity. This is the case whether or not the assumptions Quastler made in his researches were justified; in short, whether Quastler’s work was bad science or not. Nevertheless, Quastler was not reluctant to comment on the creative potential of the stochastic approach to score generation that he had made instrumental use of in the lab. He remarks:

It is amazing how much ‘character’ random music acquires by imposition of the simplest rules. Random music is quite fascinating and might be a useful tool in identifying some of the basis of musical ‘style’, its value being that it follows exactly the rules that were consciously determined and only those. We hope that it will be taken up for the sake of music, not only here, as one particular method to explore human information processing capabilities.44

Recalling that Hiller and Isaacson started work on the Illiac Suite in 1955, it is hard to believe that Quastler was not aware of interest in using the Control System Laboratory’s own computer to generate music at or around the same time as Quastler was using information-theoretic measures to quantify the musical talents of his experimental subjects. Quastler collaborated with members of the music department to design his piano sight-reading experiments; it is implausible that they would not have related details of their faculty colleague Hiller’s work to him. In any case, Quastler’s hopes, as tentative and as careful as they were, were already being fulfilled by a roster of American and European composers who turned to the principles of stochastic music over the decades that followed. This much is well-known to musicology. What to make of such apparently pre-emptory developments as Quastler’s—sometimes at the center, but so often at the margins, of twentieth-century laboratory science?

Works cited

Aspray, William. “The Scientific Conceptualization of Information: A Survey.” IEEE Ann. Hist. Comput. 7, no. 2 (April 1985): 117–40. doi:10.1109/MAHC.1985.10018.

Attneave, Fred. “Stochastic Composition Processes.” The Journal of Aesthetics and Art Criticism 17, no. 4 (1959): 503–10. doi:10.2307/428223.

Dolan, Emily I. “Musicology in the Garden.” Representations 132, no. 1 (2015): 88–94.

Feit, Anna Maria, and Antti Oulasvirta. “PianoText: Transferring Musical Expertise to Text Entry.” In CHI ’13 Extended Abstracts on Human Factors in Computing Systems on - CHI EA ’13, 3043–6. Paris, France: ACM Press, 2013. doi:10.1145/2468356.2479606.

Hiller, Lejaren, and L. Isaacson. Experimental Music: Composition with an Electronic Computer. New York: McGraw-Hill, 1959.

“ILLIAC and ORDVAC – Illinois Distributed Museum.” Accessed January 4, 2019.

Kay, Lily E. Who Wrote the Book of Life? : A History of the Genetic Code. Stanford, Calif: Stanford University Press, 2000.

Kember, Sarah. Cyberfeminism and Artificial Life. Abingdon, UK: Taylor & Francis, 2003. doi:10.4324/9780203299159.

Lennon, Brian. Passwords: Philology, Security, Authentication. Cambridge, Massachusetts ; London, England: The Belknap Press of Harvard University Press, 2018.

Quastler, Henry. “Three Survey Papers: 1) A Survey of Work Done by the Bio-Systems Group of the Control Systems Laboratory; 2) Studies of Human Channel Capacity; 3) the Informational Limitations of Decision Making.” Urbana, Illinois: Control Systems Laboratory, Univeristy of Illnois, 1956.

Quastler, Henry, and V. J. Wulf. “Human Performance in Information Transmission: Part I: General Remarks; and Part II: Sequential Tasks (Overlearned Activities).” Urbana, Illinois: Control Systems Laboratory, University of Illinois, March 1955.

Rogers, E. M. “Claude Shannon’s Cryptography Research During World War II and the Mathematical Theory of Communication.” In Proceedings 28th International Carnahan Conference on Security Technology, 1–5, 1994.

Shannon, Claude Elwood, and Warren Weaver. The Mathematical Theory of Communication. Urbana: University of Illinois Press, 1949.

  1. Claude Elwood Shannon and Warren Weaver, The Mathematical Theory of Communication (Urbana: University of Illinois Press, 1949), 3.

  2. Henry Quastler and V. J. Wulf, “Human Performance in Information Transmission: Part I: General Remarks; and Part II: Sequential Tasks (Overlearned Activities)” (Urbana, Illinois: Control Systems Laboratory, University of Illinois, March 1955),; Henry Quastler, “Three Survey Papers: 1) A Survey of Work Done by the Bio-Systems Group of the Control Systems Laboratory; 2) Studies of Human Channel Capacity; 3) the Informational Limitations of Decision Making” (Urbana, Illinois: Control Systems Laboratory, Univeristy of Illnois, 1956),

  3. Quastler, “Three Survey Papers,” 14.

  4. Brian Lennon, Passwords: Philology, Security, Authentication (Cambridge, Massachusetts ; London, England: The Belknap Press of Harvard University Press, 2018).

  5. Anna Maria Feit and Antti Oulasvirta, “PianoText: Transferring Musical Expertise to Text Entry,” in CHI ’13 Extended Abstracts on Human Factors in Computing Systems on - CHI EA ’13 (CHI ’13 Extended Abstracts on Human Factors in Computing Systems, Paris, France: ACM Press, 2013), 3044, doi:10.1145/2468356.2479606.

  6. Quastler and Wulf, “Human Performance in Information Transmission: Part I and II,” 62–18.

  7. William Aspray, “The Scientific Conceptualization of Information: A Survey,” IEEE Ann. Hist. Comput. 7, no. 2 (April 1985): 122, doi:10.1109/MAHC.1985.10018.

  8. Weaver’s essay first appeared in print as an article in Scientific American. It was prepared at the request of Chester Barnard, then the president of the Rockefeller foundation. E. M. Rogers, “Claude Shannon’s Cryptography Research During World War II and the Mathematical Theory of Communication,” in Proceedings 28th International Carnahan Conference on Security Technology, 1994, 1–5, p. 3

  9. Shannon preferred the more neutral term “output” in the earlier paper to describe the results of such communicative “procedures”.

  10. Weaver, (1964 [1949]), pp. 3–4. Weaver’s multiple returns to musical examples has perhaps not yet been noted or accounted for.

  11. Aspray, “The Scientific Conceptualization of Information,” 117.

  12. Quastler, “Three Survey Papers.”

  13. Lily E. Kay, Who Wrote the Book of Life? : A History of the Genetic Code (Stanford, Calif: Stanford University Press, 2000), p. 116. For more on Quastler’s biography, see ibid., p. 115–127.

  14. Sarah Kember, Cyberfeminism and Artificial Life (Abingdon, UK: Taylor & Francis, 2003), 17, doi:10.4324/9780203299159.

  15. Quastler, “Three Survey Papers,” 4.

  16. Ibid.

  17. Quastler and Wulf, “Human Performance in Information Transmission: Part I and II,” 62–63.

  18. Ibid., 62–65.

  19. Ibid., 4.

  20. Ibid., 6.

  21. Ibid.

  22. Ibid., 15. The vast majority of these reports were digitally scanned and made available online in the last five years.

  23. Ibid., 14.

  24. Ibid., 3.

  25. Ibid., 14.

  26. Ibid.

  27. Quastler acknowledges Stanley Fletcher (Fletcher–Munson equal loudness curves), Burrill Phillips (a composer), and Ludwig Zirner of the Music Department at the University of Illinois for “their generous help and advice.” Quastler and Wulf, “Human Performance in Information Transmission: Part I and II.”, 62-21 fn.

  28. “Three Survey Papers.”

  29. “Stochastic Composition Processes,” The Journal of Aesthetics and Art Criticism 17, no. 4 (1959): 503–10, doi:10.2307/428223.

  30. Quastler and Wulf, “Human Performance in Information Transmission: Part I and II,” 62–18.

  31. Ibid.

  32. Ibid.

  33. Ibid., 62–61.

  34. Ibid.

  35. Ibid.

  36. Ibid., 61–63.

  37. Ibid., 64.

  38. Ibid.

  39. Ibid.

  40. Emily I. Dolan, “Musicology in the Garden,” Representations 132, no. 1 (2015): 88–94,

  41. “ILLIAC and ORDVAC – Illinois Distributed Museum,” accessed January 4, 2019,

  42. Ibid.

  43. Lejaren Hiller and L. Isaacson, Experimental Music: Composition with an Electronic Computer (New York: McGraw-Hill, 1959), 5,

  44. Quastler, “Three Survey Papers,” XX.