Information and Cognitive Process: communication research for design

Peter K. Storkerson M.F.A. Ph.D.Southern Illinois University

www.communicationcognition.comemail: pstork@tiac.net telephone-fax: 716 884 0579

Information and Cognitive Process: communication research for design

Key words: communication, mode, interpretation, cognition, perception

Abstract:At the heart of the design process itself is a lacuna. Design lacks both a direct language for discuss-ing meaning as experienced and constructed by receivers, and operational methods for measuringmeaning and relating it to the design decisions that determine the spatial and temporal compositionof communications.

This paper describes a theoretical and research oriented approach to bridge the gap. The modelis based on perception, thought, and memory: species wide competencies that underlie the manifoldsocial and cultural code systems of communicative forms. Theory derived variables are exploredthrough experimental research using cross-mode communications employing visualand textual modes of presentation. The research demonstrates the efficacy of the approach, byyielding clear, dispositive and consistent results of direct relevance to designers. It demonstrates thatcommunications function by inducing and guiding the cognitive processes of receivers through themanipulation of the spatial, temporal, and behavioral characteristics of communications. As a result,important aspects of user-constructed meaning are grounded by species-wide cognitive processesand metaphors based on embodied experience.

The research protocol provides a robust, repeatable, in situ method that can be used in differentcommunication contexts for exploring highly varied communication questions and for extendingempirically substantiated theory. The research method can support the development of a larger ana-lytic base for communication design and can potentially replace survey and focus group analysiswith a tool that is simpler to use, more direct in its measures, easier to interpret, and which can beunobtrusively embedded, in situ, into pre-existing communications.

1. IntroductionLet us begin with two simple premises. First, an important part of what distinguishes communicationdesign from art and other forms of expression is its accountability: i.e. that it is required to commu-nicate something predetermined and specific to receivers. Second, the communication of contents(data or subject matter) are dependent upon frames for interpreting them. Without interpretiveframes, the contents are unintelligible, meaningless, not likely to be remembered and often not evenperceived. The dependence of communication on the receiver’s interpretive frames presents the fun-damental chicken-egg conundrum: the whole is used to interpret its parts while the parts are neededto construct the whole.

Often communicators can presume that competent receivers already possess the appropriate frames.Let us take financial communications as an example. For informed readers, the presentation of acompany’s history of earnings and its current price earnings ratio will cue a frame of referencewithin which to look at the stock price. But, we can also quickly imagine three scenarios in whichthe data display model is not adequate because it disregards the receiver.• Scenario 1: the frame of reference is unknown or problematical. The information is an enigma

that cannot be placed into any known frames. It needs the creation of a new interpretive frameto make it intelligible.

• Scenario 2: the reader is not a professional, but a consumer who needs to be educated as well asinformed. The communication must communicate both information and the frame for interpret-ing it.

• Scenario 3: the arena of decision/action is substantially different from the domains of the infor-mation. For example, the reader is presented information about a specific stock or sector, but heor she interprets the information to make decisions about personal spending, and retirementgoals and strategies.

Communication designers increasingly confront the scenarios above, reflecting the new informationrequirements of changing markets, customization of information for individual consumers who areincreasingly acting as their own fiduciaries, and particularly in terms of delivering information andproviding interfaces that make information directly actionable by end users. These challenges arerepeated in almost every area of communication. New approaches range from consolidating infor-mation, to smart agents, to qualitative (low-resolution) displays.

2. Cognitive Process ModelThese considerations bring us to an aspect of communication design that complements the notion ofcommunication as information display or as message: a cognitive process model of communication.Basic tenets of this cognitive process model can be summarized as follows.1. The content of a communication is comprised under the receiver’s formation of a concept: an

interpretive frame or frames of reference, which make the communication intelligible.2. The communicative content is not itself expressed information or subject matter.3. The content is created by the receiver using his or her cognitive faculties.4. Competent communication is the guiding of cognition.5. Cognition is a rational, human, species wide competence that both senders and receivers share.

This competence is a constant, which underlies varied cultural and social beliefs. It is also im-

plicit in communications, as a specific class of artifacts, which are created and received not asobjects in themselves, but as vehicles of communication.

6. Reliance on cognitive processes, whenever possible, will help to increase the reliability of com-munications, and their cross-cultural portability insofar as interpretation or meaning is con-cerned.

The cognitive process model of communication proposed here sees a communication not as thetransmission of information, but as a series of challenges, rewards, and affordances (resources andconstraints) that provoke and guide the receiver’s perception, thought, inference, recognition, andmemory. This model complements information presentation or message approaches by concentratingon the receiver as the one who constructs the interpretive frame. It recognizes the receiver as anactive participant. It balances this by recognizing the communication as an active participant aswell, offering the affordances that enable the receiver to achieve a usable and phenomenologicallysatisfactory interpretation. The receiver knows the content on some level and knows that he or sheknows the content.

The communication is not merely a tool, but an interactive partner, indicating to the receiver thesituation of use, and modeling the roles of sender and receiver by displaying a comportment that isappropriate to the situation, and by steering the receiver toward the frames it wishes the receiver toinvoke. When we speak of this process in terms of the document, we speak of its meaning, whilewhen we speak of the reader’s participation, we speak of interpretation. These two are inseparable,the two sides of one coin. The focus on receiver participation stresses interpretation, reflecting thespecific agendas of communication designers.

The term “interpretation” can be taken in three ways: identification, comprehension, and evaluation.Identification is the sort of interpretation being considered here. Of the three, it is primary and ob-jective or outer: it is who or what we are seeing or what is happening. It is also qualitative, as orga-nization of integrations and segregations that renders experience as distinct entities. Comprehensiongrows out of the inner consideration of identified objects and events, and evaluation is one’s senseof the objects as related to the situation and to one’s situated goals and desires. Identification is ahighly objective reasoning process. It is our orientation to the outside world. We count on it to keepus from walking into doors and falling down stairs. Communicators count on the consistency of thisinterpretive process among receivers, because they use it to anticipate, encourage, and accommodateit with appropriate affordances.

At the heart of this model is the notion that, in order to make a communication intelligible, a re-ceiver must select or create an interpretive frame that identifies content. That selection or creation isnon-trivial. It involves a cognitive leap from what is given, small or large. The frame defines thecommunication’s meaning or content, as distinct from its subject matter, and it is the first communi-cation goal: the prerequisite upon which comprehension and evaluation rest. Again, the subjectmatter may be the price of a car, while the content may be “Should I buy this car if I want to retirein comfort in fifteen years?” The enframing that yields such content can be easily cued by a spousewho whispers, “You know, dear, this one costs twice as much as any of the others.”

Increasingly, communications are combining symbolic modes such as language, with sensory modes:visualizations including charts and diagrams, sounds, motion, user actions, etc. In doing so, they

mix two different human cognitive systems. The distinction can be demonstrated directly. Take thesentence “The house fell on my head,” and ask how we make sense of it. In terms of word meanings,it is difficult to find linkages between “house” and “head” that enable the meaning to be generatedlinguistically, but it is very easy to make up a little mental visualization and watch the house riseup, rotate, turn upside down and fall on my head (Waltz, 1981). It is not so easy to visualize a sym-bolic calculation like the old chestnut “All men are mortal. If A. is a man. Then, A. is mortal.” Wecould imagine A. as a man on his death bed, but that is a different matter. It describes a specificexperience that can be interpreted within that frame

Figure 1: Mixed Model of Cognition (after Farah, 1989, p. 60)

Sensory and symbolic cognitive systems are complementary and distinct (Figure 1). Sensory systemstell us about experience. They operate according to the experiential dimensions of time and space.They function narratively, like the house visualization, and are tied to specific events and locations.Sensory systems spontaneously and involuntarily organize the flow of experience, into the discreteevents and entities. Symbol systems, on the other hand, are category driven and deployed deliber-ately. They require the naming of events and entities according to conceptual criteria. They construein terms of general categories that are not tied to any such specific places or events . They applythose categories to experience by attribution. Symbol systems concentrate on a finite number ofitems, calculating and inferring answers on the basis of reasons rather than some overall coherence.

Sensory modes give us experiential knowledge, which is incommensurable to symbolic knowledge.The house visualization does not just give more information about the house that fell on my head; ittells us how the events described might be experienced or how they might actually work. This dis-tinction parallels the one between theory and experiment or application. Theory is by and large astatic representation about relations between entities, while experiment is an experiential applica-tion, which demonstrates how theoretical relations work in a specific situation. American philoso-

phers following Willard Quine know this as an epistemological fundamental: the matching of osten-sion to category.

Figure 2: Human Cognitive Process

In graphic and multimedia communications, the symbolic and sensory modes must be integrated togive a single experiential-conceptual amalgam. This cognitive integration gives multimedia commu-nication its power. Cognitive processes combine perception and thought to form concepts or infer-ences, which are projected or imagined and compared with direct perception (Figure 2). When pro-jected inferences and direct perceptions match, the result is recognition, which is a combination ofsensory and conceptual elements equivalent to the matching of ostension to a general term or cate-gory. This matching results in a phenomenology or sense of knowing or competently grasping (Fig-ure 3) something in its entirety.

Figure 3: Cognitive Phenomenology

Here is an example. I see someone who walking toward me who strikes me as vaguely familiar. Idon’t remember her offhand, so I rack my brain to think of whom she looks like. I retrieve someonefrom the past by visualizing that person in comparison to the person in front of me. I now recognizeher and isolate the resemblance, e.g., eyes or hair. Now, recollections and feelings about that oldacquaintance also enter my mind in a flood. I owe her fifty dollars from ten years ago. Should I ap-proach her, apologize and pay it, or duck into the nearest doorway?

Graphic and multimedia communications harness the knowledge built into sensory cognition andapply it to non-sensory domains. A stock price graph relies on symbol systems to refer to days andto indices, but it is sensory cognition - the ability to run movies in the head- that is behind the in-ferred motion of diagonal lines and that enables viewers to intuit that prices are “rising” and “fal-ling”. To put it bluntly, we learn about the stock market by intuitively using our previous, sense-based understandings of climbing hills and throwing rocks. Thus, graphic communications leveragesensory knowledge and use it to teach about non-sensory concepts, and to imply the narratives theymay generate.

Aaron Sloman (Sloman, 2001. pp. 7-28) demonstrates that the combining of symbolic and sensorymodes can be seen generally in conceptualizations. The primary intuitive understanding of a prob-lem is often a sensory imagining. It is universalized by a propositional judgment that the sensoryimagining can be applied to an infinitely large class of situations. To use a trivial example, I takefour sticks, and add two. I separate the new pile into two equal ones. I take five sticks and add two,but I am not able to separate this resulting pile into two equal groups. I redo the experiment withinitial piles of six and seven sticks with the results that we know. I posit that all even numberedpiles will act alike and all odd numbered piles will act alike. So I make an inductive rule to univer-salize the observation that I intuit: If I add two to any even number, the resultant will be even, andif I do so to any odd number, the resultant will be odd. The intuitive understanding points directly to

an essential cognitive process. In this way diagrams, which are multi mode, represent a general classof reasoning as well as representation. Perhaps that is part of why they are so powerful.

Figure 4: Image-Text Cross Mode Cognition

Image-text composites can also be fascinating. Consider the photograph of Figure 4. Cover all butthe left hand image. By itself, this image has little discernible meaning. Now, cover everything butthe sentence in the middle: “Joanne used the press as the press used Joanne.” By itself, this sentence,too, is enigmatic. Now uncover the image on the right. When the text is placed next to the image asa caption, or over it, the two are combined. We have a new interpretive frame. It’s a picture of Jo-anne taken by a press photographer whom Joanne is fending-off by shielding her face. She is somesort of celebrity and the press are “using” her by invading her privacy. But, if she is also using thepress, her pose is also ironic. She is, perhaps, both fending-off and attracting the press. She is nego-tiating her celebrity. We cannot “know” these things in the veridical sense of warranted proof—thepicture could be staged and the caption was certainly added—but the combined image and text con-vey a meaning with a combination of affordances that support only certain interpretive frames, andthat afford us the phenomenology or sense of knowing what we are seeing. Phenomenologicalknowing is a primary goal of communication design. That intuitive sense of grasping makes infor-mation actionable. Such composite presentations are all the more persuasive because as receivers, wethink we are making our own interpretations.

3. OperationalizationOnce defined, cognitive processes and outputs can be measured as memory. Psychologists havedemonstrated that memory is not a record of stimuli, but of cognitive activity: its work and prod-ucts. We remember the things that we make sense of, and that we see as significant in terms of someinterpretive frame. We remember them as we interpret them: in the forms of being and meaning theyhave for us. We do not, for example, remember details in noise and clutter; we do remember con-figurations with form and structure. Memory can be seen in three aspects: retention or the fact ofremembering, comprehension which is the depth of processing that is reflected in the entities re-membered, and interpretation which is the structure or organization of associations comprising whatis remembered.

Given that memory tracks cognitive work and its results, the primary cognitive act is integration: thestable organization of the perceptual field into distinct events and entities (figure 5). Successful inte-gration supports storage in long-term memory and results in the phenomenological knowing which

can be measured as confidence. The cognitive work involved can be measured as processing time orlatency. The primary record of cognitive process is long-term retention as recall (Who was the presi-dent during the civil war?), or recognition (Who is in this picture?). Comprehension can be under-stood as the conceptual level or synthesis of memory. While verbal reports are not reliable repre-sentations of interpretation, Ericsson and Simon have demonstrated that verbalizations are limitedby the level of comprehension in a way that makes them reliable indicators of comprehension (Er-icsson and Simon, 1996). Finally, interpretation can be measured as the selectivity and structure ofmemory cues (what part reminds a receiver of what other part), which echo the interpretation.

Figure 5: Cognition and Memory

4. Experiments and FindingsThis model makes an ambitious claim to bridge from cognitive processes to interpretation. It effec-tively replaces interpretation as verbal reports, which are famously unreliable, with functional com-petencies: “If I know what you remember of a communication, I know how you can think about it,and I know what you think it is (interpretation) without your telling me.”.

Experiments can help determine whether this claim can be sustained. Two experiments were devisedfor this purpose. Both used a computer program to show movies and ask questions about them. Eachmovie was eight to twelve seconds long with a single event on video and a spoken text with a singlestatement or description (Figure 6). The experiments were entirely automated with no need for su-pervision. They were conducted in lunchrooms, libraries and airports. Data was automatically col-lected by computer, and statistical analyses utilized standard methods implemented under SPSS.

Figure 6: Experimental Movies and Questions

In the first experiment, 40 movies were shown individually to 120 subjects, both male and female,from 18 to 60 years of age and with varying levels of education. The relations between the modes(video and spoken text) varied according to whether the video and words explicitly presented com-mon subject matter (four Russian sentry guards walk as speaker discusses how they guard thebuilding), or concepts (text discussing state security), if they implicitly presented common concepts(text discussing economic disarray), or if there were no credible common concepts or subject matterpresented (text discussing flower arranging). Interpretation was operationalized as the integration ofvideo and words into a whole. Processing, or cognitive effort was measured by latency: the length oftime required to either make the integration or decide that it was not appropriate to do so. Phe-nomenological knowing was conditionally operationalized as reported confidence.

Thus, after each movie, subjects were asked whether the video and words made sense together (yesor no). Then, subjects were asked how confident they were of the judgment. Responses and responsetimes were recorded. From the data, movie scores were constructed for each movie. A movie’s inte-gration score could range from 0% if all subjects reported segregation to100% if all subjects reportedintegration. Its confidence score could range from 1 if all subjects reported low confidence to 3 if allsubjects reported high confidence. Latencies were measured in milliseconds.

Movie scores showed normal distributions with rates of integration varying from approximately 10%to 90%, confidence scores ranging from approximately 1.8 (low-medium confidence) to 2.8 (highconfidence) and latencies ranging from .2 sec. to 10 sec (the maximum allowed by the program). Themean score for integration of movies was 44%, which means that on the whole movies were inte-grated 44% of the time. On average, it took almost 2 seconds to respond to the integration question,about 1 second of which was keying and reaction time. The mean score for confidence was about2.4, indicating that most responses were in the moderate to high confidence range. The confidencelatencies averaged near one second, indicating that confidence was most likely not a consideredjudgment but an affect already aroused by the processing of the movie that could be quickly re-ported.

By definition, movies with integration scores near 50% were movies on which subjects splitas to whether they were judged to be integrated. In itself, this could be a matter of cultural or indi-vidual differences in interpretation, but other measures indicate a quite different account. Integrationlatency showed a significant U-shaped relationship to integration score (Figure 7 left), indicating

that the movies that were most often either integrated or segregated were processed most quickly.Those with scores near 50% took substantially longer to process on average, indicating that theywere more difficult to interpret for most subjects. Thus, the relationship between integration scoreand latency indicates that behind the apparent differences in interpretation for movies with scoresnear 50%, there was a common cognitive reaction. This was a strong relationship with an Rsq.greater than 40%, significant at the .000 level, indicating that subjects acted quite uniformly.

Figure7: Integration, Cognitive Difficulty, and Confidence

The movies that were difficult to process were also rated ambiguous. Subjects were most confidentof their judgment with movies scoring high and low on rates of integration (Figure 7 right). Theirconfidence dropped as the integration scores approached 50%. This was also a very strong relation-ship with an Rsq. of 61%, significant at the .000 level. Absolute confidence ratings are not in them-selves significant, reflecting, for example, a social aversion to admitting uncertainty. Here it isshapes of the curves and the uniformity of the responses that is telling

Finally, background information was collected including age, gender and educational attainment.There were some differences between groups: small differences in rates of integration, slightly differ-ent mean reported confidence levels, etc., but the U-shaped curves indicating relations between vari-ables were the same for all groups. There was no evidence of other factors.

The second experiment focused on the effects of cognitive disturbances on interpretation and mem-ory. This experiment used 20 of the movies which were used in experiment one, but here, the timerelations between video and spoken words were altered yielding four so-called delay states betweenmodes including synchronized presentation (identical to experiment one), and with either video orspoken words delayed: by one second, with one second overlap, with no overlap, or with a one sec-ond gap between whichever mode was first and the mode presented second.

As in the first experiment, immediately after each movie was shown, subjects were asked if videoand words made sense together, and to report their confidence in their judgment (low, medium, orhigh). Also, as in experiment 1, latencies were measured (Figure 6).

Figure 8: Delay, Precedence, and Integration

The Integration and Confidence results of experiment 2 showed the inhibiting effect of delays onintegration (Figure 8). Integration was highest in the synchronous 0 delay, with a score of 52%. Itdropped to as low as 35% for 1 second overlap and for no overlap. It dropped significantly, to 43%,with only a 1 second delay. Figure 8 also shows that it made little or no difference which mode wasdelayed, indicating that neither mode carried the predominance of meaning independently.

Since there was no one-to-one correspondence between videos and words, a delay of one modemerely altered already somewhat arbitrary adjacencies of words and video. It appears that subjectswere attempting to mentally synchronize modes based on perceived onset as a cue. They were tryingto realign video and words to a zero delay state using sensory or working memory. The temporal“misalignment” of modes was perceived as such because it violated the expectation that things thatbelong together start together. The lesser effect of a 1 second delay was consistent with compro-mised cognitive function as delays approach the limits of perceptual memory (1 to 1.5 seconds).

Figure 9: Recognition Test; matching video and text

In the second part of experiment 2, subjects were tested on their memory via recognition.After all of the movies were run, there was a five minute tutorial. Then, each movie was re-run,presenting one of its two modes—either the video or the spoken words—while presenting the othermode from four movies including its match (figure 9). Subjects were asked to correctly match videoand words as the movie was running. In this way, either video or spoken words could serve as a cuefor the recollection for the other, and it might be possible to detect which movies were remembered.

The overall rate of correct identification from memory was: 86%, compared to a chance 25% correctresponse rate. The overall Memory Latency, i.e. the time it took to match video and words was 8.1seconds. This reflected the difficulty of matching video and words, which involved memory anddiscrimination between memories, while presentations were being shown. The mean confidence was2.55, indicating that subjects were moderately to very confident (in relative terms) of their recollec-tion. Given the high level of correct answers, most of the calls were probably comparatively easy.

Figure 10: Memory, Delay, and Integration

The major single finding with respect to memory was that the movies that subjects had integrated inthe first part of the experiment were substantially better remembered than movies that they had notintegrated (Figure 10). Only 7% of integrated movies were not remembered while 13% of not inte-grated movies were not remembered. For integrated movies, the delay had no significant relation-ship to memory(regardless of which mode has precedence), but for segregated movies, delays im-proved memory from 72% scores for synchronized presentations to as high as 85% for delay statesof one second. As figure 10 shows, memory was highest for overlapping or gaps between modes.

This finding points toward a second factor beyond integration affecting memory: i.e., cognitive ef-fort itself. Research on “intratask interference”, challenges like the one presented by temporal shifts,indicates that such interference may inhibit initial learning, but may also actually “facilitate” longercognitive effects like memory (Battig, 1966, 1972). Under situations of interference, less will belearned but what is learned is retained longer Interference and facilitation effects are generally be-yond the scope of this paper.

5. ConclusionThis paper has proposed a cognitive processing model of communication in order to bridge the gapbetween the physical configuration of communications and received meaning. The goal of thismodel is to make a path by which meaning can be operationalized, and tested against physical vari-ables. It accomplished the goal by relating meaning to cognitive processes with measurable indica-tors.

The experiments cited are formative, but they present a strong case for the validity, researchabilityand potential usefulness of the cognitive process model. They demonstrate the effects of cognitiveprocesses that are measurable and consistent across a broad population, and that those processes arerelated to interpretation (integration), phenomenological knowing (confidence) and memory (recog-nition). The use of time delays indicates the potential importance of sensory manipulations in theinhibition or facilitation of integration and memory as well as the organization of memory.

These experiments can initiate a larger research program focusing on many other hypotheses andvariables. The method can be refined and extended in many ways:• Refinement of variables: Specific parameters like the lengths of movies or the use of images and

written texts could be altered for comparison and for refinement.• New Variables: The effects of photographic variables, sound, or music can be examined.• Cross cultural research: Socio-cultural differences affecting cognitive reception of communica-

tions can be studied.• Sequence Testing: Researching associations between movies within a sequence and between se-

quences can moving toward a comprehensive analysis of complex, extended communications.• Human-Computer Interaction: Behavioral modeling of activities and the programmed represen-

tation, elicitation and steering of human motivation (affect control) can be explored and under-stood.

• In situ testing: Experiments can be conducted outside the laboratory, in real-world situationsand integrated into design processes.

In short, the approach presented in this paper has direct implications to communication design on 3levels: findings, theory, and meta-theory: i.e., a theoretical base that can be used to produce theoriesthat are relevant to design practice.• Findings: It generates findings in the domain of practice: practical advice for communication

designers. Some findings may be surprising, such as the facilitating effects of interference.• Meta-theory: It presents a theoretical model of communications that is testable: it can be used to

make hypotheses that can be affirmed or refuted.• Empirical methods: It builds an experimental method that can be used to generate and test new

hypotheses and in that way to refine, extend and produce new models.• Creative resources: It can be can be used creatively to uncover and explore new design possibili-

ties.

Communication design finds itself between two intellectual world views: the bottom-up empiricaland quantitative approaches of science geared toward intervention, and the top down, qualitativelycomplex, symbolic, immanent and descriptive approaches in the humanist spheres of meaning. Re-cent advances both top down and bottom up suggest that it may now be possible to span the gapbetween them. This approach demonstrates that it is possible to build models that are both qualita-tively sophisticated and meaningful in the humanist sense, and empirically researchable in the sci-entific sense. It takes a step toward establishing communication design as a theoretically informed,research-based enterprise that can specify, assess and achieve communicative outcomes.

END

References

Battig, W. [1966], “Facilitation and Interference”, in Bilodeau, E., (ed) Acquisition of Skill, AcademicPress, New York, pp.134-146

Battig, W. [1972], “Intratask Interference as a Source of Facilitation in Transfer and Retention”, inThompson, R. et al (eds) Topics in Learning and Performance, Academic Press, New York, pp.134-146

Ericsson, K.& Simon, H. [1996], Protocol Analysis: Verbal Reports as Data, MIT Press, Cambridge, pp.1-10

Farah, J. M. [1989], “Text and Pictures: a Neuropsychological Perspective”, in Mandl, H. & Levin, J.(eds) Knowledge Acquisition from Text and Pictures, North Holland, New York, pp. 59-72

Lakoff, G. & Johnson, M. [1999], Philosophy in the Flesh: the Embodied Mind and its Challenge toWestern Thought, Basic Books, New York

Sloman, A. [2001] “Diagrams in the Mind?”, in Anderson, M. et al. Diagrammatic Representation andReasoning, Springer Verlag, Berlin, pp. 7-28

Storkerson, P. [2001], Cross Mode Communication in Multimedia, Ph.D. dissertation, Institute ofDesign, Illinois Institute of Technology, Available: www.communicationcognition.org ,pstork@tiac.net (Accessed 2003, April 16)

Waltz, David. 1981.“Toward a Detailed model of Processing for Language Describing the PhysicalWorld”, in Proceedings of the Seventh International Joint Conference on Artificial Intelligence,IJCAI-81, University of British Columbia, pp. 24-28