A Virtual Presence Counter - Publications...

Mel SlaterandAnthony SteedDepartment of Computer ScienceUniversity College London

London WCIE 6BT

A Virtual Presence Counter

Ab stract

This paper describes a new measure for presence in immersive virtual environments

(VEs) that is based on data that can be unobtrusively obtained during the course of a

VE experience At different times during an experience a participant will occasionally

switch between interpreting the totality of sensory inputs as forming the VE or the

real world The number of transitions from virtual to real is counted and using some

simplifying assumptions a probabilistic Markov chain model can be constructed to

model these transitions This model can be used to estimate the equilibrium probabil-

ity of being lsquolsquopresentrsquorsquo in the VE This technique was applied in the context of an ex-

periment to assess the relationship between presence and body movement in an im-

mersive VE The movement was that required by subjects to reach out and touch

successive pieces on a three-dimensional chess board The experiment included

twenty subjects ten of whom had to reach out to touch the chess pieces (the active

group) and ten of whom only had to click a handheld mouse button (the control

group) The results revealed a signicant positive association in the active group be-

tween body movement and presence The results lend support to interaction para-

digms that are based on maximizing the match between sensory data and proprio-

ception

1 Introduct ion

Imagine that you are walking through a sunlit park perhaps admiring thetrees A particular tree is interesting and you begin to move closer to it As youget within a certain distance you suddenly realize that the tree is flatmdasha virtualcardboard cutout You recall that yoursquore actually in the laboratory wearing ahead-mounted display (HMD) and it is the middle of the night After a shorttime you carry on walking through the park looking for the exit At one mo-ment you turn your head quite fast and notice that the image lags behind yourhead movement Once again you are back in the laboratory aware of theHMD and of the cables wrapped around your legs Later having found the wayout of the park you are watching the traffic and waiting for a chance to crossthe road Just as you start to move across the road you hear a voice shoutlsquolsquoYou still here at this time of nightrsquorsquo Once again yoursquore back in the laboratoryrecalled to reality by the building superintendent

This paper offers an approach to the elicitation of presence in a virtual envi-ronment (VE) based on such transitions from the virtual to the real The goalis a measurement technique that reduces reliance on questionnaires and thatgathers information during a VE experience rather than only when it is overThe measure is applied in an experiment that is designed to explore the rela-tionship between body movement and presence The results indicate a positive

Presence Vol 9 No 5 October 2000 413ndash434

r 2000 by the Massachusetts Institute of Technology

Slater and Steed 413

association between these two which is important forthe design of interaction paradigms for immersive VEsThis study reproduces the findings of another recentexperiment on the relation between body movementand presence (Slater et al 1998) but which used tradi-tional questionnaire-based methods

Because the approach to the elicitation of presenceintroduced in this paper is quite different to anythingtried before it is worthwhile to briefly discuss the overallmethodology Based on a set of simplifying assumptionsa stochastic process is developed in an attempt to modelthe number of transitions experienced by VE partici-pants between the two states lsquolsquopresence in the VErsquorsquo andlsquolsquopresence in the real worldrsquorsquo Based on the stochasticmodel an estimator for the proportion of time spent inthe lsquolsquopresence in the VErsquorsquo state is constructed Of coursethere is no suggestion that the assumptions or the modelthat flows from it are lsquolsquotruersquorsquo as a description of real-world mental or behavioral processes rather they pro-vide an abstraction against which data can be collected

2 Back ground

In this section we consider the background ontwo interrelated questions the definition of presenceand its elicitation and measurement An extensive reviewcan be found in Draper Kaber and Usher (1998) whoidentify three types of presence in the literature simplecybernetic and experiential The first is simply the abilityto operate in the virtual environment and the second isconcerned with aspects of the human-computer inter-face In this paper we concentrate on the third ofDraper et alrsquos typesmdashthe experiential approach inwhich presence lsquolsquois a mental state in which a user feelsphysically present within the computer-mediated envi-ronmentrsquorsquo This follows the common view that presenceis the sense of lsquolsquobeing therersquorsquo in the virtual environmentor equivalently presence in a virtual environment in-volves the sense of being in the virtual place rather thanin the real physical place (such as the laboratory) wherethe personrsquos body is actually located (Held amp Durlach1992 Sheridan 1992 Barfield amp Weghorst 1993Slater Usoh amp Steed 1994 Sheridan 1996 Ellis1996 Witmer amp Singer 1998)

A quite different approach (Zahoric amp Jenison 1998)gives a working definition of presence as lsquolsquotantamount tosuccessfully supported action in the environmentrsquorsquo Thisapproach emphasizes that reality is grounded in actionrather than in mental filters and sensations and that lsquolsquothereality of experience is defined relative to functionalityrather than to appearancesrsquorsquo (Flach amp Holden 1998)This approach concentrates on action (how participantsdo things) rather than how things look and sound andthat being there is actually the ability to do there The VEbecomes endowed with lsquolsquothere-nessrsquorsquo through this pro-cess of action and interaction The present paper sup-ports this view and indeed the experiment does sup-port the idea that action in the sense of appropriatewhole-body movements is associated with a higher senseof presence It will be argued that one group of our sub-jects had the chance to learn their environment throughbody movements rather than through just seeing it andthat these subjects reported a greater sense of presenceNevertheless we do see value in a construct such as pres-encemdashthe sense of being theremdashthat is independent ofthe lsquolsquoactionrsquorsquo in which it is embedded

Some authors have distinguished between immersionand presence In this view immersion is a term used fordescribing the technology that can give rise to presenceFor example Draper Kaber and Usher (1998) writethat lsquolsquoimmersion is the degree to which sensory input toall modalities is controlled by the SE synthetic environ-ment interfacersquorsquo A fundamental research goal is to un-derstand how presence is influenced by these (immer-sive) properties of the system that generates the VE andby the rules and methods by which people interactwithin it (Slater amp Wilbur 1997) Ellis (1996) has ar-gued that the equation relating presence to its influenc-ing factors must be such that iso-presence curves can beconstructed thus allowing an understanding of howdifferent configurations and combinations of these fac-tors can be used to attain a given level of presence Thisis an important engineering requirement allowingtradeoffs between various system components for ex-ample for economic considerations Bystrom Barfieldand Hendrix (1999) have recently produced a frame-work for presence research that also embodies this dis-tinction between immersion and presence

414 PRESENCE VOLUME 9 NUMBER 5

Despite the widespread agreement on the concept ofpresence its measurement has many different ap-proaches Some studies avoid the issue of measuringpresence by examining the impact of more (or less) pres-ence on success in the performance of a task within aVE For example Pausch Proffit and Williams (1997)used this approach in the context of a search task Pres-ence was not measured but was assumed to follow froma more immersive setup (use of a head-tracked head-mounted display) However what this study reallyshowed was the impact of different levels of what wehave called immersion on the task performance ratherthan presencemdashfor a simple relationship between im-mersion and presence cannot be assumed The need tomeasure presence independently of immersion is neces-sary also because immersion might influence presenceand task performance in different ways

The qualitative or ethnographic approach to presenceinvolves in-depth studies with relatively few people orwith case histories in order to gain substantial insightinto the phenomenon and its relationship to other fac-tors This approach is exemplified for example by thework of McGreevy (1993) and Gilkey and Weisenberger(1995) Both examined the notion of presence amongpeople in relatively extreme circumstances the first ofgeologists in the field on a Mars-like terrain and undervarying arrangements of their visual field of view andthe second of the suddenly deafened adult Profoundinsights into the nature of presence (or its absence) canbe elicited in this way

Another approach is to attempt to measure presenceby observing peoplersquos behavior Held and Durlach(1992) suggested a lsquolsquostartlersquorsquo or looming response Thiswas extended by Sheridan (1992) to lsquolsquosocially condi-tionedrsquorsquo responses (would a person involuntarily put outtheir hand in response to a hand-shake gesture) InSlater Usoh and Chrysanthou (1995) an attempt wasmade to measure presence behaviorally by introducingcontradictory information about an object representedin both the real and virtual world with some informa-tion (visual) coming from the VE and other information(auditory) from the real world The extent to which par-ticipants respond to the visual information (allowing fordifferences in sensory preference) indicates their degree

of presence in the virtual A similar approach althoughin the vestibular domain has been tried by Prothero etal (1995)

These behavioral techniques all suffer from the sameproblem some feature or task has to be added to theenvironment (to cause the looming response for ex-ample) that may have nothing to do with the applica-tion but is only there for the purpose of measuring pres-ence A good property of a measuring instrument shouldbe the extent to which it can be used in any applicationwithout the addition of particular features that are forthe sole purpose of measurement

A variation on the behavioral approach was suggestedby Barfield and Weghorst (1993) and by Welch (1997)Both studies suggest that presence in the VE would in-duce aftereffects in participants once they had left theVE The intensity and extent of such aftereffects couldbe used as a measure of presence As Welch points outthough the relationship is a complex one He arguesthat the intensity of such aftereffects will have a negativecorrelation with the degree of initial presence and apositive correlation with longer-term adaptive presenceAlthough measurement techniques could be constructedon this basis none have appeared in the literature todate

By far the most common approach to the elicitation ofthe degree of presence is through subjective reportingusually in conjunction with a questionnaire Barfield andWeghorst (1993) measured presence using subjectivereporting on ten-point scales with three questions thatproved to be highly correlated the sense of being therethe sense of inclusion in the virtual world and the senseof presence in the virtual world Slater Steed and Usoh(1993) required subjects to rate their sense of lsquolsquobeingtherersquorsquo in the virtual environment the extent to whichthe VE became their dominant reality and the extent towhich the VE became a place rather than just imagesEach of these was rated on a scale of 1 to 7 and thepresence score was taken as the number of high scores(6 or 7) This questionnaire has been extended in severalfurther studies adding more questions each based onone of the three main ideas most recently in Usoh et al(1999)

Welch et al (1996) used the method of paired com-


parisons Presence was defined to the subjects and theywere required to choose a value between 1 and 100 toindicate the lsquolsquosize of perceived differencersquorsquo between thepresence-impact of two environments Their definitionof presence lsquolsquoemphasized the feeling that subjects werephysically located in and surrounded by the portrayedvisual world rather than in the laboratory in which theyknew the experiment to be taking placersquorsquo

Hendrix and Barfield (1996a) used a questionnairewith an anchored 1-to-100 scale around three aspects ofpresence one question involved use of the term presenceitself the second the sense of being there in the virtualworld and the third the level of realism of the virtualworld Subjects were instructed to answer 100 for asense of presence equivalent to the real world and 1 forno presence The same approach was used in their laterstudy (Hendrix amp Barfield 1996b)

Witmer and Singer (1998) define presence as lsquolsquothe sub-jective experience of being in one place or environmenteven when one is physically situated in anotherrsquorsquo andlsquolsquopresence refers to experiencing the computer-gener-ated environment rather than the actual physical localersquorsquoThey construct a questionnaire based on 32 influencingfactors each rated on a scale of 1 to 7 such as controlfactors (lsquolsquoHow much were you able to control eventsrsquorsquo)sensory factors (lsquolsquoHow much did the visual aspects of theenvironment involve yoursquorsquo) distraction factors (lsquolsquoHowaware were you of events occurring in the real worldaround yoursquorsquo) and realism factors (lsquolsquoHow inconsistentor disconnected was the information coming from yourvarious sensesrsquorsquo) Their presence measure is then thesum of responses to the 32 questions and they also tryto identify important determinants by examining corre-lations between the individual factors and the overallsum This approach is interesting because unlike theearlier-mentioned approaches their measure does notdirectly attempt to elicit presence according to theirdefinition of the word Instead the measure is based onsubjective responses to various aspects of immersion (inthe sense described earlier in this paper as properties ofthe VE delivering system itself) A critique of Witmerand Singerrsquos approach may be found in Slater (1999)followed by their reply

One problem with the use of subjective rating scales is

that the scores are ordinal and therefore strictly shouldnot be combined together to form summations TheSlater Steed and Usoh (1993) approach avoids thisproblem by taking the overall score as a count of highresponses out of the number of presence questions Un-der the null hypothesis that responses are statisticallyindependent the overall count has a binomial distribu-tion and therefore logistic regression may be used inanalysis Snow and Williges (1998) obtain a ratio-scalemeasurement of presence by using free modulus magni-tude estimation Each subject provides a number (de-gree of presence) in response to a stimuli and is free togive a different number in response to different stimuliSnow and Williges show how to combine these numbersto obtain a ratio-scale that is valid within and betweensubjects They seem to side-step the issue of definingpresence and report that subjects lsquolsquowere asked to assigna number to their feeling of how much they felt as ifthey were actually present in the virtual environmentduring performance of the tasks in that trialrsquorsquo

Still using the same idea of subjective reporting Free-man et al (1999) adapt methods of continuous assess-ment of TV-picture quality to assessing presence Whileviewing a VE subjects are able to manipulate a handheldslider indicating their sense of presence lsquolsquodefined forobservers as lsquoa sense of being therersquo in a displayed sceneor environmentrsquorsquo The experiments involved changingvarious aspects of the presentation (such as from mono-scopic to stereo) and indeed corresponding changes inthe recorded presence scores were found

A difficulty with this procedure is that there is no con-trol When observers see some change in the displayquality the only response available to them in the con-text of the experiment is either to do nothing or to movethe slider As a control it would be interesting to repeatthe series of experiments described but calling the slidera measure of lsquolsquofactor Xrsquorsquo and examining whether the pat-tern of responses is similar to those of presence

Freeman et al also show how subjective assessmentcan be influenced by the prior training and experimentalconditions Although this is a general problem in pres-ence research it is exacerbated by the tendency to usewithin-group experimental designs in which subjectsexperience a number of different stimuli and are asked to


rate the presence level corresponding to each (whetherduring or after the trial) When studies on humans in-volve responses that are involuntary (for example pupildilation in response to brightness) it is reasonable andnormal to use within-group designs However in thecase of presence where the response is subjective andvoluntary subjects can clearly be influenced in their re-sponses by the information that they gather during thecourse of the experiment For example if they experi-ence two VEs one with full color and the other onlymonochrome they can quickly figure out that the ex-perimenter (other things being equal) would be expect-ing a higher sense of presence in the full-color one Inthe context of the slider this problem is taken to an ex-treme because subjects experience within the same ses-sion different aspects of the quality of the display so thatit must become very obvious what the point of the ex-periment is and what the expectations associated with itare Freeman et al are for this reason aiming to movetowards more-objective methodologies

From the discussion above it would be easy to con-clude that presence is a relationship of an individual toan environment and that the degree of presence isquantifiable and may vary more or less continuously Weconsider these assumptions in more detail Imagine anindividual receiving and aware of sensory stimuli fromonly one environment1 In our approach to presencethe individual would by definition be present in thatenvironment The issue of presence becomes interestingonly when there are competing environmentsmdashthat isthe individual is receiving and can be aware of stimulifrom multiple environments (including internal ones)Presence then determines which of the environments theindividual responds to and acts within at any given mo-ment Slater Usoh and Steed (1994) defined the dis-played environment (ED) as that created by the VE sys-tem Then given any environment E the notationp(E ED) was used to represent the degree of presence inE given that the individual was lsquolsquoinrsquorsquo ED Presence in theVE was then defined as p(ED ED)

A rigorous analysis by Schloerb (1995) operational-ized this by treating the degree of presence explicitly as aprobability measure lsquolsquothe degree of subjective presenceis defined to be the probability that the person perceivesthat he or she is physically present in the given environ-mentrsquorsquo Schloerb describes a thought experiment inwhich an individual randomly receives stimuli from thephysical environment or from a similar virtual environ-ment and each time is asked to state which of the twoenvironments (physical or virtual) he or she is in Thisallows a probability estimation of the degree of presence

As Draper Kaber and Usher (1998) have pointedout Schloerbrsquos approach treats subjective presence asbinary where experience is lsquolsquobifurcated into telepresentand not-telepresent experiences whereas other authorsconsider telepresence continuously scalablersquorsquo

If we return to the idea of presence as the potential toact in an environment then all this can be tied togetherSuppose the individual is receiving stimuli from a num-ber of competing environments (the physical world inwhich the person is standing the virtual world and in-ternal mental worlds such as memories and day-dreams) At any moment the person responds and actswith respect to one of these environments By this wemean both the shifting of attentional resources towardsthe specific signals belonging to that environment aswell as the response to those signals direction of gazeattention to particular sounds awareness of parts of thebody such as the pressure on the soles of the feet causedby standing on the floor awareness of a sudden draftand voluntary actions such as body movements and ut-terances Returning to the opening example of this pa-per at one moment the individual is aware of and re-sponding to the stimuli from the virtual park and is notpaying attention to the vast array of other signals thatconstitute the physical environment of the laboratoryand its surroundings When presence in the park is bro-ken for example by a glitch in the display or a suddennoise from the real world at that moment the individualis hardly aware of the virtual park stimuli and muchmore aware of the temperature in the laboratory theweight of the HMD sensations such as pressures andcontact with other objects on his or her physical bodyand so on We can think of presence as a selector among

1 Here we include internal stimuli as constituents of valid internalmental environments so the worlds of daydreaming would be includedas an environment


environments to which to respond which operates dy-namically from moment to moment If it were possibleto freeze time at a specific instant the individual wouldbe paying attention and responding to a set of stimulithat corresponds to one environment and ignoringstimuli from other environments Also they may be in-terpreting stimuli from one environment in the contextof the currently present one (for example interpreting asound from the real world as belonging to the virtualworld) A fundamental proposal of this paper is that theset of stimuli of the present environment forms an overallgestalt providing a consistent believable world in itselfThis overall idea of considering attentional resources aspart of the determination of presence is also included inthe Immersion Presence and Performance frameworkof Bystrom Barfield and Hendrix (1999)

Now the view of presence as the selection of oneamong multiple environments at any moment in time iscertainly not inconsistent with continuous measures Forexample at any moment in an experience the individualcan be asked to rate the degree of presence in the VEand the response is determined by an integration overthe last small interval of time Or at the end of an ex-periment a questionnaire rating may be determined byfor example an integration over the whole time periodSchloerbrsquos thought experiment of asking the individualto choose between presence in the VE and presence inthe real world is also compatible with this the individualwould say lsquolsquoyesrsquorsquo to presence in the VE if at that momentshe or he was about to respond to the VE stimuli ratherthan the real-world physical or other stimuli

The main contribution of this paper is an attempt tointroduce a measure that is this integration over time Itis based on the idea of a gestalt formed from stimuli atany moment in time and the construction of a verysimple stochastic model for this Within the context ofthe stochastic model there is a parameter (presence overtime) that has an unknown value and this paper pro-vides a method to estimate this The estimation methodrelies on the number of transitions from presence in theVE to presence in the real worldmdashwhere these transi-tions are reported during the course of the VE experi-ence The method is far from ideal but is to our knowl-edge the first attempt along these lines

The new technique is introduced in the next section

It is applied in an experiment in Section 4 The purposeof the experiment was two-fold to assess whether thenew measure gives results that are comparable to theusual questionnaire results and to examine a hypothesisthat the degree of body movement in a VE task is posi-tively associated with presence The results are given inSection 5 with discussion and critique in Section 6 Theconclusions and some recommendations for future usesof the method are presented in Section 7

3 A Presence Counter

3 1 Introduct ion

Gestalt psychology (Kohler 1959) contains a no-tion of figure and ground within a single figure (Figure1a) (Perls Hefferline amp Goodman 1969) one aspectmight come to the foreground thus giving one interpre-tation or another aspect might come to the foregroundresulting in a quite different interpretation Just as tran-sitions occur between figure and ground in gestalt psy-chology so in VE experiences people often report suchtransitions between the real and the virtual

Figure 1 Gestalt images


Figure 1a through 1d are examples of the well-knownresult that the same information can be perceptually in-terpreted as quite different entities by the same person atdifferent moments in time Figure 1c for example willusually be first interpreted as three narrow triangularsectors radiating from the center of the circle How-ever after staring at the center for a while the figurewill suddenly reorganize itself into something differentand then every so often a spontaneous change fromone interpretation to the other will occur (Kohler1959)

While in an immersive VE the participant receives acontinuous stream of sensory datamdashmainly visual fromthe VE but also often auditory from the real world andof course real-world tactile and kinesthetic data (such asthe weight of the helmet) Occasionally the VE sensorydata exhibits glitches such as when the frame rate sud-denly changes (for example a more complex part of thescene comes into view) or when a close-up view of anobject reveals its texture mapping Occasionally real-world data will intrude a telephone rings there is a sud-den movement of air as a door is opened the tempera-ture changes a cable wraps around a leg Sometimesinternal mental processes of the participants will sparkthe realization that they are actually in a VE or wearinga head-mounted display really in some laboratory orexhibition hall and not in the illusory place presented tothem by the VE

In other words at any moment two alternate gestaltsare available to the individual experiencing a VE state V(lsquolsquoI am in the place depicted by the VE systemrsquorsquo) andstate R (lsquolsquoI am in a lab in the Computer Science build-ing wearing a helmetrsquorsquo) At each moment the indi-vidual will tend towards one rather than the other Pres-ence in the VEmdashvirtual presencemdashmay be thought of asthe extent to which the interpretation V is favored

During the course of a VE experience however assuggested an individual will typically experience transi-tions between states V and R The moments at whichthe individual switches from one interpretation to theothermdashin particular from V to Rmdashare of particular in-terest If it could be known when and why these oc-curred this would be a major contribution to the prob-lem of eliciting the factors that enhance or inhibit virtual

presence The participants cannot be asked to reporttransitions from R to V because this would require themto immediately break out of their state of presence toreport back to the real world However it is a postulateof this paper that they can be asked to report transitionsfrom V to R

The last point is controversial and it is argued by anal-ogy with common experience Suppose you were askedas is usual with many systems of meditation to quietyour mind to stop conscious thought Many readerswould have tried this Now suppose that an additionalinstruction were given At the moment you haveachieved a quiet mind report this to your meditationinstructor This is impossible for obvious reasons How-ever suppose the additional instruction were instead asfollows You may achieve a quiet mind but if at anytime you become aware that thoughts are once againbuzzing through your head please report this to yourinstructor Now it is possible to achieve a quiet mindand while in such a state the instruction to report atransition to an active mind is not in awareness (other-wise a quiet mind would not have been achieved) It isreasonable to assume that when you become aware ofconscious thoughts again you will remember to reportthis transition to the instructor

Another example is again a common experience be-coming very absorbed in a movie While so absorbedyou are typically oblivious to your real surroundingseven oblivious to the state of your body Every so oftenthough some real world event or some event within themovie itself will occur that will throw you out of thisstate of absorption and back to the real world of the the-ater someone nearby unwraps a candy bar someonecoughs some aspect of the storyline becomes especiallyridiculous and so on The reporting of transitions intothe state of absorption is impossible without undermin-ing the absorbed state itself However reporting transi-tions back to reality are obviously possible

Now virtual presence is being structurally likened tothe quiet mind or the absorbed state in a movie Whilethe participant is virtually present there is no logicalrequirement that they must be thinking about reportingtransitions to R for if they are thinking this then theyare not (yet) in state V It is this strong definition of vir-tual presence that is adopted in this paper


A further analogy can be found in the study ofdreams A researcher in a dream research laboratoryknows the likely onset of a dream by observing rapid eyemovement (REM) At any moment during the REMphase the sleeper can be awoken and asked to report thedream2 In the case of the VE experience if the state ofpresence is considered equivalent to a dream thedreamer is awakened by whatever caused the break inpresence At that moment a report can be given that abreak has occurred without this in itself disturbing thesense of presence which of course has already been dis-turbed

3 2 A Stochast ic Model for Break sin Presence

Consider the following scenario An individual en-ters a VE with the instruction to report whenever a breakin presence (BIP) occurs and to report this only at sucha moment At the end of the experience lasting time tthere will be b such BIPs at times t1 t2 tb Theproblem now is to use this information to recover thetendency (p) of the individual to be in the presence stateand also to understand the reasons why the BIPs oc-curred when they did Here p is given a specific interpre-tation as the asymptotic (long-term equilibrium) prob-ability of being in state V This is not particularlydifferent from Schloerbrsquos interpretation of subjectivepresence although the method of estimation is different

It is clearly difficult to recover p from the time se-quence because only half the information is available(that is when and how many times there was a break inpresence is known) but the times when the individualentered the presence state are unknown When b 5 0(no transitions) for example is this because the indi-vidual spent the whole time present (in state V) or thewhole time in state R For any given value of b there aretwo extreme possible interpretations one in which theunknown time is assumed to be in the V state and theother in which the unknown time is assumed to be in theR state The discrepancy between these two interpreta-tions decreases with increasing b Assuming that the

transitions from R to V and V to R occur instanta-neously at random moments in time according to a Pois-son process it is easy to show that the expected value ofp for increasing b is 05 (Appendix A)

An estimator for p can be constructed with some sim-plifying assumptions Suppose that presence is binarythat is at any moment of time the participant is either instate R or state V Discretize time by dividing the totaltime into n equal intervals (t 5 1 2 n) Denote bypij the probability that if at time interval t the participantis in state i they will be in state j at the next time intervalt 1 1 (that is a BIP has occurred on the boundary be-tween these two time intervals) Here state 0 corre-sponds to R and state 1 to V Note that this assumes pij

to be independent of t so the transition matrix

P 5p00 p01

p10 p11(EQ 1)

represents a stochastic process modeled by a two-stateMarkov chain (Karlin 1969)

It is not difficult to show that P k is the k-step transi-tion matrix its elements P ij

(k) are the probabilities that ifat time t the individual is in state i then at time t 1 kthey will be in state j As k ` the equilibrium prob-abilities p0 and p1 are obtained denoting the probabili-ties of being in the corresponding states in the long run(which given the assumptions of a Markov chain areindependent of the initial state) A fundamental limittheorem of Markov chains shows that (in the particularcase of the two-state chain)

p0 5 p0p00 1 p1p10

p1 5 p0p01 1 p1p11

p0 1 p1 5 1

(EQ 2)

and therefore

p0 5p10

p01 1 p10

p1 5p01

p01 1 p10

(EQ 3)

The unknown p is interpreted as p1 the equilibriumprobability of being in state V

2 An excellent popular account of this type of research can befound in S LaBergersquos Lucid Dreaming (Ballantine Books 1985)


The goal now is to use the observed data t1 t2 tb

to estimate the transition probabilities pij under each oftwo alternate conditions the first assuming a low pro-pensity to presence and the second a high propensity(bearing in mind the two possible interpretations ofb 5 0) In each case the ti are assigned to the appropri-ate intervals and mark a transition from state 1(V) to0(R) such transitions are assumed to occur at theboundary between the two intervals

321 Low-Presence Condition There are bobserved transitions from V to R (BIPs) If t is an inter-val at which there were such a transition then at inter-val t 2 1 the participant must have been in state VThere are therefore b intervals with state V It is as-sumed for the moment that intervals are small enoughso that no two successive states reported a BIP and thatthe first and last intervals did not report a BIP In thelow-presence condition it is assumed that all intervals inwhich the state is unknown are in state 0

p00 is the proportion of times that an interval in state 0is followed by an interval also in state 0 There are n 2

1 2 b intervals in state 0 that are followed by anotherinterval (Interval n has no successor) All but b of themare followed by intervals in state 0 Therefore

p00 5n 2 1 2 2b

n 2 1 2 b

p01 5b

n 2 1 2 b

(EQ 4)

where 2b n 2 1p10 is the proportion of times that an interval in state 1

is followed by an interval in state 0 Because this alwaysoccurs

p10 5 1 and p11 5 0 (EQ 5)

From these the equilibrium probabilities are

p0 5n 2 1 2 b

n 2 1(EQ 6)

and

p1 5b

n 2 1

where 2b n 2 1

322 High-Presence Condition In this casethe assumption is that all intervals in which the state isunknown are in the state V(1) A similar analysis to thatabove yields

p0 5b

n 2 1(EQ 7)

and

p1 5n 2 1 2 b

n 2 1

where 2b n 2 1Let pc(b) be the equilibrium probability of being in

state V with b BIPs observed and with c correspondingto the low-presence (L) condition or the high-presence(H) condition Then

pL(b) 5b

n 2 1(EQ 8)

and

pH (b) 5n 2 1 2 b

n 2 1

where 2b n 2 1The relationship is illustrated in Figure 2 showing

that when the number of BIPs achieves its maximum(n 2 1)2 p 5 05

Figure 2 highlights a problem In practice only b isobserved and for any level of b there are two extremevalues of p Knowing only b gives insufficient informa-tion to estimate p To choose between pL(b) and pH (b)

Figure 2 Relationship between number of BIPs (b) and overall

presence (p)


therefore a discriminator is requiredmdashsome additionalinformation to select one of the two alternatives Thesimplest way to achieve this is a question at the end ofthe session asking the participant to classify their overallexperience with respect to their sense of presence Theanswer to this together with the value of b would thenallow an estimate of p

It should be recalled that this analysis gives for anycondition two extreme interpretations For example inthe low-presence condition the analysis implies that

b

n 2 1 p

1

2 (EQ 9)

In the absence of prior information it would be normalto use an estimate halfway between these two boundsSuch estimates would be linear transforms of pL(b) andpH (b) and therefore would have no effect on relation-ships with other variables discovered in statistical analy-sis Therefore this paper continues to use pL(b) andpH (b) which should properly be referenced as lsquolsquoextremalprobabilitiesrsquorsquo although the qualifier lsquolsquoextremalrsquorsquo is usu-ally dropped

323 Special Cases For any choice of time in-terval there can always be a situation in which successiveintervals report a BIP or in which there is a BIP in thefirst interval or in the last interval It is important to beable to cater for these special cases in order to avoid theproblem of having to choose very large values of n thusforcing the probabilities to the extremes

The analysis can be easily adjusted to take this intoaccount One additional assumption is made If there aresuccessive BIPs then the amount of time in the V statein-between them is negligible

Suppose that k out of the b BIPs are followed by a BIPin the next interval Then the transition matrix prob-abilities are as follows

P(low 2 presence)

5

n 2 1 2 2(b 2 k)

n 2 1 2 b 1 k

b 2 k

n 2 1 2 b 1 k

1 0(EQ 10)

with

pL(b) 5b 2 k

n 2 1(EQ 11)

and

P(high 2 presence)

5

k

b

b 2 k

b

b 2 k

n 2 1 2 b

n 2 1 2 2b 1 k

n 2 1 2 b

(EQ 12)

with

PH (b) 5n 2 1 2 b

n 2 1 (EQ 13)

A further refinement allows for a BIP in the first or lastintervals Let s1 5 1 if there is a BIP in the first intervaland 0 otherwise and similarly sn 5 1 if there is a BIP inthe last interval and 0 otherwise Then following thesame reasoning above it can be shown that the transi-tion matrix probabilities are

P(low 2 presence)

5

n 2 1 2 2(b 2 s1 2 k)

n 2 1 2 (b 2 s1) 1 k

b 2 s1 2 k

n 2 1 2 (b 2 s1) 1 k

1 0

(EQ 14)

and

P (high 2 presence)

5

k

b 2 sn

b 2 sn 2 k

b 2 sn

b 2 s1 2 k

n 2 1 2 (b 2 sn)

n 2 1 1 k 2 2b 1 s1 1 snn 2 1 2 (b 2 sn)

(EQ 15)

Clearly b k and when b 5 0 then k 5 0 s1 5 sn 5 0and the probabilities would be 0 or 1 in this case

Figure 3 illustrates the ideas of this section It shows


the occurrences of the BIPs and the corresponding lowand high extremal probabilities (multiplied by 10) for aparticular individual selected from the subjects in theexperiment described in Section 4 The BIPs are taken asoccurring on the boundaries between intervals but theyactually occur within an interval This is illustrated in thegraph by the steeply sloping lines in an interval in whicha BIP has occurred

4 Experiment

This section describes an experiment that had twoprincipal goals First the above technique is used to esti-mate p If the approach is sound then there should be asignificant positive correlation between p and postexperi-mental questionnaire results relating to presence follow-ing the hypothesis that these questions are answered onthe basis of the balance of time that a participant spendsin the V state compared to the R state Second previousresults (for example Slater et al (1998) and Usoh et al(1999)) suggest a positive relationship between the de-gree of body movement of participants and their re-ported presence The experiment also examines this ideain a different context to those previous studies

The experimental scenario involved the participantsobserving a sequence of moves on a three-dimensionalchess board (as introduced by the Star Trek TV series)which was chosen because it is a quite large and complexthree-dimensional object and fitted well with the re-quirement to induce significant body movement in par-ticipants who were required to reach out and touch thechess pieces A stereo pair of the three-dimensional chessboard is shown in Figure 4 As can be seen it is a struc-ture with several layers resting on a table The dimen-sions are shown in Table 1

The pieces that had to be touched were distributedover the entire board Hence the highest piece was ap-proximately 146m above the ground which for somesubjects required considerable stretching to reach It isimportant to note that the measure of body movementused (namely the total amount of hand movement) is of

Figure 3 Atime sequence showing occurrences of BIPs and the

corresponding probabilities for one subject

Figure 4 A stereo pair showing the 3-D chess model

Table 1 Three-Dimensional Chess Dimensions

ObjectDimension(m)

Table top 074Large chess boards 02 3 02Small chess boards 01 3 01Lowest board height above table top 022Middle board height above table top 042Highest board height above table top 062Small boards height above large board 01


course a measure of whole body movement because itwould encompass such reaching and stretching

The factorial design was for twenty participants di-vided into two groups as shown in Table 2

Each subject was paid pound5 (about $9) for completingthe study and were recruited through advertisements onthe campus In the event two subjects were unable toeither understand or properly follow the instructions andwere replaced by two other subjects so 22 people com-pleted the experiment with two cases being discardedThe final twenty comprised eighteen men and twowomen five undergraduates six Masters students fourPhD students two research assistants one faculty mem-ber and two miscellaneous others No subject had anyknowledge of or anything at all to do with the researchitself

Each participant started the experience in a virtuallaboratory (the virtual anteroom) After receiving in-structions the participants made their way through adoor to a field with trees and plants outside Some fivemeters beyond the door was a table with the 3-D chessboard (Figure 5 and 6) Those assigned to the Low Ac-tivity group were told to repeatedly look for a red chesspiece and when it was found to press a button on a 3Dmouse that they were holding throughout and to ob-serve the movement of the piece Those assigned to theHigh Activity group were told that when they observedthe red piece to reach out with their hand (holding the3-D mouse) and touch it and it would then move Atthe end of an entire sequence of nine moves a large but-ton on the side of the virtual table would turn red TheLow Activity group had to click the physical button ontheir handheld 3-D mouse and the High Activity grouphad to reach and touch the virtual red button All par-ticipants were told that when they noticed that the skyhad become dark they should return from the field to

the starting room The sky was darkened after threecomplete sequences of moves and the mean and stan-dard deviation of the time spent in the field was319 sec 6 64 sec All subjects were told that they wouldbe asked about the sequence of moves observed after theexperience

Prior to starting the experiment all subjects wereasked to complete a short questionnaire that obtained

Table 2 Factorial Design

Lowactivity

Highactivity

Number of subjects 10 10

Figure 5 View from within the virtual lab out to the eld

Figure 6 Overview of the lab area


background information gender job status and priorexperience of virtual reality

After completing the questionnaire the subject wasshown each of Figure 1b through 1d in turn (Becausemost subjects had seen Figure 1b before it was hardlyused) They were asked to describe their initial interpre-tation of the figure (For example for Figure 1c mostsaw the three thin triangles first) Then they were askedto stare at the figure and notice if any change occurredIf a change did happen they were asked to continue toobserve the figure and clearly exclaim lsquolsquoNowrsquorsquo if andwhenever it spontaneously reconfigured itself to look thesame way as when they first saw it After this trainingthey were given the instructions in Table 3 to read

All instructions were reinforced verbally once the sub-ject entered the real laboratory itself and then againwhile they were in the virtual anteroom before enteringthe field that contained the chess board While in theanteroom they were shown how to move around howto make a small red cube on a table respond by eithertouching it (High Activity group) or by clicking withtheir forefinger on the 3-D mouse (Low Activity group)

The virtual reality laboratory is in a small enclosedroom within a large laboratory in which there is con-tinual noise (constant noise of workstations and randomnoise of occasional phone rings or conversations) Noattempt was made to reduce background noise or to fur-ther isolate the VR room from the remainder of thelaboratory indeed there was interest as to whether the

background events would trigger transitions from Vto R

The scenarios were implemented on a Silicon GraphicsOnyx with twin 196 MHz R 10000 Infinite RealityGraphics with 64M main memory The software usedwas Divisionrsquos dVS and dVISE 312 The tracking sys-tem has two Polhemus FASTRAKs one for the HMDand another for a five-button 3-D mouse The helmetwas a Virtual Research VR4 which has a resolution of742 3 230 pixels for each eye 170660 color elementsand a field of view of 67 deg diagonal at 85 overlap

The total scene consisted of 13298 polygons runningat a frame rate of no less than 20 Hz in stereo The la-tency was approximately 120 ms

Subjects moved through the environment in gaze di-rection at constant velocity by pressing a thumb buttonon the 3-D mouse They had a simple inverse kinematicvirtual body (Figure 7) When they reached forward totouch a chess piece they would see their virtual arm andhand

At the end of the session subjects were given a secondquestionnaire the main purpose of which was to gatherinformation on their sense of presence An initial ques-tion asked for the reason why (if this was the case) theyreported no or very few transitions giving four optionsrarely being in the virtual world almost always being in

Table 3 Transition-to-Reality Instructions

IMPORTANT Transition to Reality

When you enter the virtual reality you may have thesense of being in lsquolsquoanother placersquorsquo

In just the same way as you saw transitions in theimages that you just looked at you may experiencetransitions in your sense of placemdash

Virtual sometimes you will be in the virtual placeReal sometimes you will become aware of the real lab

in which the experience is really happeningIf and only whenever you experience a transition to Real

please say lsquolsquoNowrsquorsquo very clearly and distinctively

Figure 7 Exocentric view of someone reaching out to touch a chess

piece


the virtual world forgetting to report transitions orother reasons In retrospect this question was not par-ticularly useful because it required an answer only whensubjects reported no or very few transitions withoutgiving a definition of this No subject reported forget-ting the instruction to report transitions

A second question was open-ended asking for thecauses of the transitions (whether or not these had beenreported at the time) Five questions related to presencewere interspersed throughout the questionnaire eachrated on a scale of 1 to 7 where 1 indicated low pres-ence and 7 high presence These questions followed thesame model introduced by Slater Steed and Usoh(1993) as briefly discussed in Section 2 The first ques-tion was a priori considered the most direct elicitation ofpresence and used as the discriminator a score of morethan 4 on this resulted in the formula pH (b) being usedotherwise pL(b)

The five questions relating to presence were

1 Please rate your sense of being in the field on thefollowing scale from 1 to 7 where 7 representsyour normal experience of being in a placeI had a sense of lsquolsquobeing therersquorsquo in the field(1) Not at all (7) Very much

2 To what extent were there times during the experi-ence when the field became the lsquolsquorealityrsquorsquo for youand you almost forgot about the lsquolsquoreal worldrsquorsquo ofthe laboratory in which the whole experience wasreally taking placeThere were times during the experience when the vir-tual field became more real for me compared to thelsquolsquoreal worldrsquorsquo (1) At no time (7) Almost all the time

3 When you think back about your experience doyou think of the field more as images that you sawor more as somewhere that you visited Please an-swer on the following 1 to 7 scaleThe virtual field seems to me to be more like (1) Images that I saw (7) Somewhere that I vis-ited

4 During the time of the experience which wasstrongest on the whole your sense of being in the

field or of being in the real world of the labora-toryI had a stronger sense of being in (1) The real world of the laboratory (7) The virtualreality of the field of plants

5 During the time of the experience did you oftenthink to yourself that you were actually just stand-ing in an office wearing a helmet or did the fieldoverwhelm youDuring the experience I often thought that I wasreally standing in the lab wearing a helmet (1) Most of the time I realized I was in the lab (7)Never because the virtual field overwhelmed me

Some data were automatically collected during thecourse of the experiment in particular the times atwhich the participant said lsquolsquoNowrsquorsquo and the total time inthe virtual field The amount of hand and head move-ment was computed by the program running the simula-tion as a function of the head and hand tracking

5 R esu lt s

5 1 General

The overall levels of reported presence as ascer-tained from the questionnaire responses were high Fig-ure 8 shows the median response for each of the fivepresence-related questions showing for example thaton question 1 (the discriminator) half of the responseswere at level 5 or higher

The number of BIPs ranged between 0 and 14 Themean time between BIPs was 48 sec 6 37 sec theminimum time interval was 55 sec and the maximum141 sec The time interval used for the analysis was 10sec this being approximately the largest compatiblewith the assumptions that 2b n 2 1 There were twocases where there were some BIPs in sequence and twoother cases where there was a BIP in the first or last in-terval

The questionnaire-based presence using all five pres-ence questions is plotted against the number of BIPS inFigure 9


5 2 Relat ionship b et w een p andQuest ionnaire Based Presence

The first question to consider is whether a relation-ship exists between the estimate of presence p and thequestionnaire responses The usual approach of the au-thors to combining the results of the presence questionsinto one overall score (without resorting to averagingacross ordinal data) is to count the number of high

scores (6 or 7) thus giving each subject a count out of4 for the questions other than that used as the discrimi-nator (questions 2 to 5 above)

Figure 10 shows the scatter plot of p against the ques-tionnaire presence count (recall that the prsquos are extre-mal) There is a clear positive relationship although withone outlying point Even including this point there is astatistically significant correlation between p and thepresence count (r2 5 032 t 5 2920 t18 5 2101 atthe 5-significance level) When this outlier is removedthe result improves substantially (r2 5 065 t 5 5588t17 5 3965 at 01) Examining the responses of theparticular person represented by the outlier he wrotethat he was disturbed by the absence of sound in the VEknew that the experimenters were in the real lab along-side him and that he wanted to talk to them becauseexploring an environment is often a lsquolsquocommunal activ-ityrsquorsquo The experimenterrsquos notes record that he did indeedcontinue to talk to them during the immersive experi-ence He gave a response of 3 to the discriminator ques-tion (writing lsquolsquoSOUNDrsquorsquo next to his response) andscores of 7 for each of the remaining four presence-re-lated questions The assignment of this person to thelow-presence condition on the basis of this particulardiscriminator is therefore dubious

Figure 8 Median levels of reported presence

Figure 9 Number of BIPs by Questionnaire Presence

Figure 10 Scatter plot of p against Questionnaire Presence


5 3 R elat ionsh ip betw een p and HandAct iv ity

The next issue to consider is the relationship be-tween p and the main independent factors Table 4shows the means and standard deviations of p for theactivity groups and the difference in means is not sig-nificant although contrary to expectation the Low Ac-tivity group seems at first sight to have a higher averagepresence than the High Activity group There is a highlysignificant difference in variance (the variance ratio is140 F99 5 32 at 5) with much less variation amongthe Low Activity group The difference in variation wasto be expected because the subjects in the inactivegroup were not required to move their hands at all (ex-cept to press the button on the handheld 3-D mouse fornavigating which in fact did not require any movementof the hand relative to the body)

A more detailed examination reveals a different situa-tion showing that the differences in means cannot betaken at face value Figure 11 shows a scatter plot of p bytotal hand movement per unit time discriminating be-tween the two activity groups It suggests that there isno discernible relationship between the amount of handmovement of the Low Activity group and p as would beexpected However for the High Activity group there isa positive linear relationship between hand movementand p Therefore comparing raw means for the twogroups not taking into account the hand movement isanyway invalid

Table 5 gives the result of a multiple regression analy-sis of p on activity and hand movement (HM) allowingfor the possibility of differing slopes for the Low Activityand High Activity groups There is a significant positiveslope for the High Activity group with an overallsquared multiple correlation of 038

Inspection of Figure 11 shows an outlying point witha low presence and high hand movement (above 015mps) in the High Activity group This was caused by thesame person who was the outlier in Figure 10 Remov-ing the data for this person from the analysis results inTable 6 The squared multiple correlation increases to073 and the slope for the High Activity group is wellinto the highly significant range

531 Explanations for BIPs One questionasked the participants to give the reasons for their transi-

Table 4 Means and Standard Deviations for p by Activity

Mean and standard deviation

Activity High 060 6 044 (n 5 10)Activity Low 089 6 012 (n 5 10)

t 5 199(t16 5 2120 at 5)

Figure 11 Presence against hand movement

Table 5 Regression of P on Activity and Hand Movement perUnit Time (HM)

Activity Regression t for slope

High p 5 2 091 1 1107 HM(se 5 490)

226

Low p 5 087 1 034 HM(se 5 478)

007 (ns)

Allows for different slope for HM for each level ofactivity Based on n 5 20 observations with df 5 16t16 5 2120 at 5 Overall R2 5 038


tions to the real If you did make transitions from virtualto real whether or not you reported these at the time whatdo you remember as the causes of the transitions (For ex-ample hearing an unexpected noise from the real labmight cause such a transition)

The reasons given can be classified into two maintypes

ExternalmdashSensory information from the real world in-truded into or contradicted the virtual world either inthe form of noises or people talking or else the touchor feel of interactions with real solid objects (such asthe VR equipment itself)

InternalmdashThis is where something lsquolsquowrongrsquorsquo with thevirtual world itself is noticed such as the laws of phys-ics not being obeyed objects looking unreal the ab-sence of sounds or display lag

There were a number of subsidiary reasons

ExperimentmdashSome aspect of the experimental set-upitself or the instructions intruded

PersonalmdashSome personal feeling intruding such as em-barrassment or consciousness of being observed fromthe outside

AttentionmdashA loss of attention to what is happening inthe virtual world or some aspect of the virtual worldthat results in a loss of presence

SpontaneousmdashA BIP for no (conscious) apparent rea-son

Table 7 gives the number of participants who re-sponded in each of these categories and some examplesof each

532 The Discriminator Question The analy-sis above hinges on the choice of the discriminator ques-tion because it classifies each participant into a low-pres-ence or high-presence group and therefore determinesthe computation of p A different discriminator questioncould lead to quite different results However the re-sults for this experiment are robust with respect to thechoice of discriminator question

The analysis was repeated for each of the remainingfour presence-related questions and also for the averageof all of the five questions (Table 8) (Of course the out-lying point corresponding to someone who had written3 for question 1 but 7 for each of the others does notoccur for any of the choices of discriminator other thanquestion 1) For every choice of discriminator questionexcept for question 3 the results are the same When themean response of all of the presence questions is used asdiscriminator the results are again the same (Note thatquestion 4 has exactly the same impact as a discriminatoras the average)

6 Discussion

The method presented in this paper relies on anumber of assumptions

1 Presence in the lsquolsquorealrsquorsquo and lsquolsquovirtualrsquorsquo is treatedas a binary state The authors would not seek todefend this as a statement about the psychologicalprocesses involved It is used here in the spirit of asimplifying assumption to allow the constructionof the stochastic model However arguments werepresented at the conclusion of Section 2 in favor ofthe notion that presence may be considered as aselection of one environment relative to which anindividual acts at a given moment

2 The stochastic model assumes discrete timeAgain this is a simplifying assumption that is oftenemployed in the initial stages of constructing amodel of complex phenomena It may be possibleto employ a continuous-time stochastic model in-stead

Table 6 Regression of p on Activity and HM(Outlier Removed)


High p 5 2 190 1 1911 HM(se 5 336)

569

Low p 5 087 1 034 HM(se 5 292)

011 (ns)

Based on n 5 19 observations with df 5 15 t16 5

2131 at 5 Overall R2 5 073


3 The transitions can be modeled as a Markov chainThis assumes that the transition probabilities are one-step that what happens in any interval is statisticallyindependent of all other intervals except for the lastThe veracity of this assumption is unknown and againshould be viewed as a simplification for the purpose ofconstructing an initial model

4 The requirement to report BIPs does not in it-self influence the participants to report BIPsExperimental evidence supports the argument thatthe requirement to report BIPs increases thechance of BIPs occurring Girgus Rock and Egatz(1977) found that giving subjects a knowledge ofthe reversibility of ambiguous figures substantially

Table 7 Reasons Given for Transitions to the Real

Cause n Some examples

External sound 7 lsquolsquoNoises from the lab (people talking)rsquorsquolsquolsquoHearing background noisersquorsquo

External touchor force

9 lsquolsquoI was supposed to be in a grass plain but when I moved my feet I realized it wasa plank under my feet (in the real)rsquorsquo

lsquolsquoFeeling of the floor under my feetrsquorsquolsquolsquoBecoming aware of cable wrapped around footrsquorsquolsquolsquoThe cable brushing against my legsrsquorsquolsquolsquoTrapped in wiresrsquorsquo

Internal 1 lsquolsquoThe length of time taken to interact with the worldrsquorsquo1 lsquolsquoTurning and thus becoming aware that the virtual world was not the real worldrsquorsquo

lsquolsquoIf moved head quicklyrsquorsquolsquolsquoThe time taken for the chess pieces to moversquorsquolsquolsquoThe way the sky darkened not smooth like someone had switched off the sunrsquorsquolsquolsquoWeird things happening that are obviously not real (eg the chess set)rsquorsquolsquolsquoThe more I needed to examine the contents of the lsquovirtualrsquo the more my aware-

ness flipped into the lsquorealrsquo rsquorsquolsquolsquoBecame very close to the chess boardrsquorsquolsquolsquoGetting too near to things (especially trees)rsquorsquo

Experiment 3 lsquolsquoHaving to remember the real-world instructionsrsquorsquolsquolsquoThe task of being asked to monitor the changes from virtual to reality itself cre-

ates a sense of going back to realityrsquorsquolsquolsquoOnce experienced a transition became sensitive to it happening againrsquorsquo

Personal 3 lsquolsquoWanted to talk to the experimenters to share the experiencersquorsquolsquolsquoEmbarrassmentrsquorsquolsquolsquoVery consciousrsquorsquo

Attention 3 lsquolsquoTransitions occurred between the tasks for example when looking for the nextred piece but only if I couldnrsquot see it at first glancersquorsquo

lsquolsquoAttention wandered after realizing that the chess sequence was iterativersquorsquolsquolsquoNot having a task to dorsquorsquo

Spontaneous 2 lsquolsquoSpontaneous feelingrsquorsquolsquolsquoIt just occurred to mersquorsquo

n is number of participants who gave responses in the corresponding category


increased the chance of these being reported Ap-proximately half the subjects who were not toldabout the reversibility of figures never reported atransition whereas all of the subjects who weretold about the reversibility always reported transi-tions This raises the possibility that more BIPswere reported than would otherwise have naturallyoccurred

This is a difficult issue because it could also beargued that the requirement to report BIPs sets upa dual task for the participants to do their actualtask in the VE and to pay attention to their state inorder to be able to report the BIPs A counter-argument to this is that the requirement to reporta BIP is likely to enter consciousness only at thetime immediately after a BIP has occurred (as dis-cussed earlier)

A preferable response to these problems is toagree that the method for reporting BIPsmdashrelyingon a verbal responsemdashis certainly not an ideal wayto obtain this information It is an interesting andchallenging research topic to try to find physiologi-cal correlates to BIPs that can be measured unob-trusively

5 The discriminator question can discriminatebetween the low- and high-propensity casesUse of a discriminator question does indeed resultin an uncomfortable reliance on questionnairedata An alternative behaviorally based discrimina-tor would be preferable

7 Conclusions

Notwithstanding the critique above a new methodfor measuring presence in virtual environments has beenintroduced in which the major component of the mea-sure depends on data collected during the course of theVE experience itself It is based on the number of transi-tions between the state of being in the VE to the state ofbeing in the real world Using the simplifying assump-tion that changes in state between presence and non-presence to form a time-independent Markov chain anequilibrium probability of presence can be estimatedThis requires only one additional postexperimental dis-criminator question concerning each participantrsquos assess-ment of whether they had been in the presence state formore or for less than half the time

Table 8 Regression Analysis p on Hand Movement (HM) Computing p with different discriminators Only the High Activityequations are shown (no Low Activity result is signicant) n 5 20 df 5 16 t16 5 2120

Question used as discriminator R2

Regressionp 5

t-value

1 Please rate your sense of being in the field on the following scale from 1 to 7where 7 represents your normal experience of being in a place

038 2 091 1 1107 HM 226

2 To what extent were there times during the experience when the field be-came the lsquolsquorealityrsquorsquo for you and you almost forgot about the lsquolsquoreal worldrsquorsquoof the laboratory in which the whole experience was really taking place

042 2 164 1 1700 HM 341

3 When you think back about your experience do you think of the field moreas images that you saw or more as somewhere that you visited

026 2 023 1 778 HM 146

4 During the time of the experience which was strongest on the whole yoursense of being in the field or of being in the real world of the laboratory

077 2 146 1 1623 HM 701

5 During the time of the experience did you often think to yourself that youwere actually just standing in an office wearing a helmet or did the fieldoverwhelm you

043 2 164 1 1667 HM 320

Discriminator is computed as average of responses to questions 1 through 5 077 2 146 1 1623 HM 701


This new technique was tried in an experiment to as-sess the impact of body movement on presence and itwas found to be positively and significantly correlatedwith the usual questionnaire-based measure It is en-couraging that the questionnaire score and the new mea-sure were correlated However further studies are neces-sary for validation

Another issue for the experiment apart from themethodology for presence measurement was the rela-tionship between presence and hand movement (In thisexperiment head and hand movement were significantlycorrelated r2 5 054 t 5 46 t18 5 2101 at 5) Theevidence strongly suggests a positive association betweenpresence and hand movement in line with previous evi-dence on the relationship between presence and bodymovement The direction of causality is unknown butthe authors suspect that there is a two-way relationshiphigh presence leads to greater body movement andgreater body movement reinforces high presence Thereis some evidence to support the first part of this state-ment Because subjects in the High Activity group wereall required to reach out and touch the chess pieces whyis there such high variation in their measured handmovements since they all had to reach the same dis-tances The variation can be explained by the fact thatsome subjects took the most-direct routes to the chesspieces ignoring collisions with the board and otherpieces whereas others acted more as they would have inreal life avoiding collisions with other objects Hence itcould be argued that presence in the VE caused them toact this way thus leading to greater body movement

The relationship between presence and body move-ment follows from the notion that one of the most im-portant determinants for presence is the requirement fora match between proprioception and sensory data Thisis consequential for the design of interaction paradigmswhere semantically appropriate body movement exploit-ing proprioception is preferred for example to the im-portation of techniques from 2-D interfaces This rela-tionship between presence and proprioception in thedesign of interaction paradigms has been exploited inseveral studies (Slater Usoh amp Steed 1995 MineBrooks amp Sequin 1997 Grant amp Magee 1998 Usohet al 1999)

We suggest some recommendations for future use of

the new measure First the discriminator questionshould ask for the information that is required in a verydirect manner (Question 4 would have been preferableto question 1) As soon as the VE experience has termi-nated the participant should be asked to estimate theoverall proportion of time spent in the presence statecrucially whether above or below 50 The exact word-ing of this discriminator question or indeed as noted inSection 6 whether there is some better way to obtainthis information should be given more thought

Second a standard time interval should be agreed sothat results can be easily compared between differentapplications and systems In this experiment the timeinterval was chosen to be the greatest compatible withthe requirement that 2b n 2 1 (n is the number ofintervals) The interval used was 10 sec The choice oflarge values of n grants undue weight to the statisticalsignificance of the count data and pushes the probabilityestimates to more-extreme values (although it does notalter the relationship between them)

The results are nevertheless robust with respect to therange of time intervals An analysis with intervals rang-ing from 1 sec through to the maximum compatiblewith the crucial requirement of 2b n 2 1 always givesthe same results A preferable solution would be to con-struct a model that does not require the use of discretetime intervals

Finally to return to the issue of body movement Inthe 3-D chess experiment it is clear that many of the(active) subjects are learning about the chess board withtheir whole bodies As stated earlier to call the move-ments lsquolsquohand movementsrsquorsquo is an understatement of whatis being measured It is not surprising that among theactive group those who exhibited greater body move-ment also tended to have a greater sense of being in thesame space as the 3-D chess board

Ack now ledgments

This work is funded by the European ACTS COVEN ProjectThe authors would like to thank V Linakis for the model ofthe 3-D chess board The authors would like to thank Profes-sor Frederick Brooks Jr (University of North Carolina) for hisencouragement and helpful comments


References

Barfield W amp Weghorst S (1993) The sense of presencewithin virtual environments A conceptual framework In GSalvendy and M Smith (Eds) Human Computer Interac-tion Software and Hardware Interfaces (pp 699ndash704) Am-sterdam Elsevier

Bystrom K-E Barfield W amp Hendrix C (1999) A con-ceptual model of the sense of presence in virtual environ-ments Presence Teleoperators and Virtual Environments8(2) 241ndash244

Draper J V Kaber D B amp Usher J M (1998) Telepres-ence Human Factors 40(3) 354ndash375

Ellis S R (1996) Presence of mind A reaction to ThomasSheridanrsquos lsquolsquoFurther musings on the psychophysics of pres-encersquorsquo Presence Teleoperators and Virtual Environments5(2) 247ndash259

Flach J M amp Holden J G (1998) The reality of experi-ence Gibsonrsquos way Presence Teleoperators and Virtual Envi-ronments 7(1) 90ndash95

Freeman J Avons S E Pearson D E amp IJsselstijn W A(1999) Effects of sensory information and prior experienceon direct subjective ratings of presence Presence Teleopera-tors and Virtual Environments 8(1) 1ndash13

Gilkey R H amp Weisenberger J M (1995) The sense ofpresence for the suddenly deafened adult Presence Teleop-erators and Virtual Environments 4(4) 357ndash363

Girgus J J Rock I amp Egatz R (1977) The effect ofknowledge of reversibility on the reversibility of ambiguousfigures Perception and Psychophysics 22(6) 550ndash556

Grant S C amp Magee L E (1998) Contributions of pro-prioception to navigation in virtual environments HumanFactors 40(3) 489ndash497

Held R M amp Durlach N I (1992) Telepresence PresenceTeleoperators and Virtual Environments 1 109ndash112

Hendrix C amp Barfield W (1996a) Presence within virtual envi-ronments as a function of visual display parameters PresenceTeleoperators and Virtual Environments 5(3) 274ndash289

mdashmdashmdash (1996b) The sense of presence within auditory envi-ronments Presence Teleoperators and Virtual Environments5(3) 290ndash301

Karlin S (1969) A First Course in Stochastic Processes NewYork Academic Press

Kohler W (1959) Gestalt Psychology An Introduction to NewConcepts in Modern Psychology New York Liverright

McGreevy M W (1993) The presence of field geologists in aMars-like terrain Presence Teleoperators and Virtual Envi-ronments 1(4) 375ndash403

Mine M R Brooks F P Jr amp Sequin C (1997) Movingobjects in space Exploiting proprioception in virtual-envi-ronment interaction Computer Graphics (SIGGRAPH) Pro-ceedings 19ndash26

Pausch R Proffit D amp Williams G (1997) Quantifyingimmersion in virtual reality Computer Graphics(SIGGRAPH) Annual Conference Series 13ndash18

Perls F S Hefferline R F amp Goodman P (1969) GestaltTherapy Excitement and Growth in the Human PersonalityNew York Julian Press

Prothero J Parker D Furness T amp Wells M (1995) To-wards a robust quantitative measure of presence Proceed-ings of the Conference on Experimental Analysis and Mea-surement of Situational Awareness (pp 359ndash366) DaytonaBeach FL Embry-Riddle Aeronautical University Press

Schloerb D (1995) A quantitative measure of telepresence Pres-ence Teleoperators and Virtual Environments 4(1) 64ndash80

Sheridan T B (1992) Musings on telepresence and virtualpresence Presence Teleoperators and Virtual Environments1 120ndash126

mdashmdashmdash (1996) Further musings on the psychophysics of pres-ence Presence Teleoperators and Virtual Environments 5(2)241ndash246

Slater M (1999) Measuring presence A response to the Wit-mer and Singer presence questionnaire Presence Teleopera-tors and Virtual Environments 8(5) 560ndash565

Slater M Steed A amp Usoh M (1993) The virtual tread-mill A naturalistic method for navigation in immersive vir-tual environments In Martin Goebel (Ed) First Eurograph-ics Workshop on Virtual Environments (pp 71ndash83)Polytechnical University of Catalonia

Slater M Usoh M amp Steed A (1994) Depth of presencein virtual environments Presence Teleoperators and VirtualEnvironments 3(2) 130ndash144

mdashmdashmdash (1995) Taking steps The influence of a walking meta-phor on presence in virtual reality ACM Transactions onComputer-Human Interaction (TOCHI) 2(3) 201ndash219

Slater M Usoh M amp Chrysanthou Y (1995) The Influ-ence of Dynamic Shadows on Presence in Immersive VirtualEnvironments Second Eurographics Workshop on VirtualReality M Goebel ed Monte Carlo Jan 1995

Slater M amp Wilbur S (1997) A framework for immersivevirtual environments (FIVE) Speculations on the role ofpresence in virtual environments Presence Teleoperators andVirtual Environments 6(6) 603ndash616

Slater M Steed A McCarthy J amp Maringelli F (1998)The influence of body movement on subjective presence invirtual environments Human Factors 40(3) 469ndash477


Snow M P amp Williges R C (1998) Empirical models basedon free-modulus magnitude estimation of perceived presencein virtual environments Human Factors 40(3) 386ndash402

Usoh M Arthur K Whitton M Bastos R Steed ASlater M amp Brooks F (1999) Walking Walking-in-place Flying in virtual environments Computer Graphics(SIGGRAPH) Annual Conference Series 359ndash364

Welch R B (1997) The presence of aftereffects Advances inHuman FactorsErgonomics 21A In G Salvendy M J Smithamp R J Koubek (Eds) Design of Computing Systems CognitiveConsiderations (pp 273ndash276) Elsevier Amsterdam

Welch R B Blackman T T Liu A Mellers B A amp StarkL W (1996) The effects of pictorial realism delay of visualfeedback and observer interactivity on the subjective senseof presence Presence Teleoperators and Virtual Environ-ments 5(3) 263ndash273

Witmer B G amp Singer M J (1998) Measuring presence invirtual environments A presence questionnaire PresenceTeleoperators and Virtual Environments 7(3) 225ndash240

Zahoric P amp Jenison R L (1998) Presence and being-in-the-world Presence Teleoperators and Virtual Environments7(1) 78ndash89

A ppendix A The Ex pected Time ``Present rsquo rsquoas the Numb er of BIPs Increases

Consider the transitions from R to V and V to R as oc-curring instantaneously at random moments in time ac-

cording to a Poisson process that is each time interval isan independent exponentially distributed random vari-able The result is the same and mathematically simpler ifeverything is normalized by the total time leading to 2bobservations from the uniform probability distributionon the interval 0 1 Suppose the BIPs occur at timest2 t4 t2b The times at which the transitions R to Voccur are t1 t3 t2b2 1 with 0 t1 t2 t2bTherefore the total time in the state V TV is given bythe formula

TV 5 oi 5 1

i 2 b

(t2i 2 t2i2 1) (EQ 16)

The expected value of ti is i(2b 1 1) resulting in

E(TV) 5b

2b 1 1(EQ 17)

Therefore

as b ` E(TV)1

2 (EQ 18)

In the time interval from t2b to 1 there may be anothertransition from R to V However the expectation of thisextra time in V will tend to 0 with increasing b


association between these two which is important forthe design of interaction paradigms for immersive VEsThis study reproduces the findings of another recentexperiment on the relation between body movementand presence (Slater et al 1998) but which used tradi-tional questionnaire-based methods

Because the approach to the elicitation of presenceintroduced in this paper is quite different to anythingtried before it is worthwhile to briefly discuss the overallmethodology Based on a set of simplifying assumptionsa stochastic process is developed in an attempt to modelthe number of transitions experienced by VE partici-pants between the two states lsquolsquopresence in the VErsquorsquo andlsquolsquopresence in the real worldrsquorsquo Based on the stochasticmodel an estimator for the proportion of time spent inthe lsquolsquopresence in the VErsquorsquo state is constructed Of coursethere is no suggestion that the assumptions or the modelthat flows from it are lsquolsquotruersquorsquo as a description of real-world mental or behavioral processes rather they pro-vide an abstraction against which data can be collected

2 Back ground

In this section we consider the background ontwo interrelated questions the definition of presenceand its elicitation and measurement An extensive reviewcan be found in Draper Kaber and Usher (1998) whoidentify three types of presence in the literature simplecybernetic and experiential The first is simply the abilityto operate in the virtual environment and the second isconcerned with aspects of the human-computer inter-face In this paper we concentrate on the third ofDraper et alrsquos typesmdashthe experiential approach inwhich presence lsquolsquois a mental state in which a user feelsphysically present within the computer-mediated envi-ronmentrsquorsquo This follows the common view that presenceis the sense of lsquolsquobeing therersquorsquo in the virtual environmentor equivalently presence in a virtual environment in-volves the sense of being in the virtual place rather thanin the real physical place (such as the laboratory) wherethe personrsquos body is actually located (Held amp Durlach1992 Sheridan 1992 Barfield amp Weghorst 1993Slater Usoh amp Steed 1994 Sheridan 1996 Ellis1996 Witmer amp Singer 1998)

A quite different approach (Zahoric amp Jenison 1998)gives a working definition of presence as lsquolsquotantamount tosuccessfully supported action in the environmentrsquorsquo Thisapproach emphasizes that reality is grounded in actionrather than in mental filters and sensations and that lsquolsquothereality of experience is defined relative to functionalityrather than to appearancesrsquorsquo (Flach amp Holden 1998)This approach concentrates on action (how participantsdo things) rather than how things look and sound andthat being there is actually the ability to do there The VEbecomes endowed with lsquolsquothere-nessrsquorsquo through this pro-cess of action and interaction The present paper sup-ports this view and indeed the experiment does sup-port the idea that action in the sense of appropriatewhole-body movements is associated with a higher senseof presence It will be argued that one group of our sub-jects had the chance to learn their environment throughbody movements rather than through just seeing it andthat these subjects reported a greater sense of presenceNevertheless we do see value in a construct such as pres-encemdashthe sense of being theremdashthat is independent ofthe lsquolsquoactionrsquorsquo in which it is embedded

Some authors have distinguished between immersionand presence In this view immersion is a term used fordescribing the technology that can give rise to presenceFor example Draper Kaber and Usher (1998) writethat lsquolsquoimmersion is the degree to which sensory input toall modalities is controlled by the SE synthetic environ-ment interfacersquorsquo A fundamental research goal is to un-derstand how presence is influenced by these (immer-sive) properties of the system that generates the VE andby the rules and methods by which people interactwithin it (Slater amp Wilbur 1997) Ellis (1996) has ar-gued that the equation relating presence to its influenc-ing factors must be such that iso-presence curves can beconstructed thus allowing an understanding of howdifferent configurations and combinations of these fac-tors can be used to attain a given level of presence Thisis an important engineering requirement allowingtradeoffs between various system components for ex-ample for economic considerations Bystrom Barfieldand Hendrix (1999) have recently produced a frame-work for presence research that also embodies this dis-tinction between immersion and presence

































3 1 Introduct ion




















P 5p00 p01

p10 p11(EQ 1)




p0 5 p0p00 1 p1p10

p1 5 p0p01 1 p1p11

p0 1 p1 5 1

(EQ 2)

and therefore

p0 5p10

p01 1 p10

p1 5p01

p01 1 p10

(EQ 3)









p00 5n 2 1 2 2b

n 2 1 2 b

p01 5b

n 2 1 2 b

(EQ 4)



p10 5 1 and p11 5 0 (EQ 5)


p0 5n 2 1 2 b

n 2 1(EQ 6)

and

p1 5b

n 2 1

where 2b n 2 1


p0 5b

n 2 1(EQ 7)

and

p1 5n 2 1 2 b

n 2 1



pL(b) 5b

n 2 1(EQ 8)

and

pH (b) 5n 2 1 2 b

n 2 1





presence (p)




b

n 2 1 p

1

2 (EQ 9)





P(low 2 presence)

5

n 2 1 2 2(b 2 k)

n 2 1 2 b 1 k

b 2 k

n 2 1 2 b 1 k

1 0(EQ 10)

with

pL(b) 5b 2 k

n 2 1(EQ 11)

and

P(high 2 presence)

5

k

b

b 2 k

b

b 2 k

n 2 1 2 b

n 2 1 2 2b 1 k

n 2 1 2 b

(EQ 12)

with

PH (b) 5n 2 1 2 b

n 2 1 (EQ 13)


P(low 2 presence)

5

n 2 1 2 2(b 2 s1 2 k)

n 2 1 2 (b 2 s1) 1 k

b 2 s1 2 k

n 2 1 2 (b 2 s1) 1 k

1 0

(EQ 14)

and

P (high 2 presence)

5

k

b 2 sn

b 2 sn 2 k

b 2 sn

b 2 s1 2 k

n 2 1 2 (b 2 sn)

n 2 1 1 k 2 2b 1 s1 1 snn 2 1 2 (b 2 sn)

(EQ 15)





4 Experiment








ObjectDimension(m)










Lowactivity

Highactivity






















piece












5 R esu lt s

5 1 General






















t 5 199(t16 5 2120 at 5)




High p 5 2 091 1 1107 HM(se 5 490)

226

Low p 5 087 1 034 HM(se 5 478)

007 (ns)















6 Discussion






High p 5 2 190 1 1911 HM(se 5 336)

569

Low p 5 087 1 034 HM(se 5 292)

011 (ns)


2131 at 5 Overall R2 5 073

























7 Conclusions




Regressionp 5

t-value


038 2 091 1 1107 HM 226


042 2 164 1 1700 HM 341


026 2 023 1 778 HM 146


077 2 146 1 1623 HM 701


043 2 164 1 1667 HM 320











Ack now ledgments



References








































TV 5 oi 5 1

i 2 b

(t2i 2 t2i2 1) (EQ 16)


E(TV) 5b

2b 1 1(EQ 17)

Therefore

as b ` E(TV)1

2 (EQ 18)


































3 1 Introduct ion




















P 5p00 p01

p10 p11(EQ 1)




p0 5 p0p00 1 p1p10

p1 5 p0p01 1 p1p11

p0 1 p1 5 1

(EQ 2)

and therefore

p0 5p10

p01 1 p10

p1 5p01

p01 1 p10

(EQ 3)









p00 5n 2 1 2 2b

n 2 1 2 b

p01 5b

n 2 1 2 b

(EQ 4)



p10 5 1 and p11 5 0 (EQ 5)


p0 5n 2 1 2 b

n 2 1(EQ 6)

and

p1 5b

n 2 1

where 2b n 2 1


p0 5b

n 2 1(EQ 7)

and

p1 5n 2 1 2 b

n 2 1



pL(b) 5b

n 2 1(EQ 8)

and

pH (b) 5n 2 1 2 b

n 2 1





presence (p)




b

n 2 1 p

1

2 (EQ 9)





P(low 2 presence)

5

n 2 1 2 2(b 2 k)

n 2 1 2 b 1 k

b 2 k

n 2 1 2 b 1 k

1 0(EQ 10)

with

pL(b) 5b 2 k

n 2 1(EQ 11)

and

P(high 2 presence)

5

k

b

b 2 k

b

b 2 k

n 2 1 2 b

n 2 1 2 2b 1 k

n 2 1 2 b

(EQ 12)

with

PH (b) 5n 2 1 2 b

n 2 1 (EQ 13)


P(low 2 presence)

5

n 2 1 2 2(b 2 s1 2 k)

n 2 1 2 (b 2 s1) 1 k

b 2 s1 2 k

n 2 1 2 (b 2 s1) 1 k

1 0

(EQ 14)

and

P (high 2 presence)

5

k

b 2 sn

b 2 sn 2 k

b 2 sn

b 2 s1 2 k

n 2 1 2 (b 2 sn)

n 2 1 1 k 2 2b 1 s1 1 snn 2 1 2 (b 2 sn)

(EQ 15)





4 Experiment








ObjectDimension(m)










Lowactivity

Highactivity






















piece












5 R esu lt s

5 1 General






















t 5 199(t16 5 2120 at 5)




High p 5 2 091 1 1107 HM(se 5 490)

226

Low p 5 087 1 034 HM(se 5 478)

007 (ns)















6 Discussion






High p 5 2 190 1 1911 HM(se 5 336)

569

Low p 5 087 1 034 HM(se 5 292)

011 (ns)


2131 at 5 Overall R2 5 073

























7 Conclusions




Regressionp 5

t-value


038 2 091 1 1107 HM 226


042 2 164 1 1700 HM 341


026 2 023 1 778 HM 146


077 2 146 1 1623 HM 701


043 2 164 1 1667 HM 320











Ack now ledgments



References








































TV 5 oi 5 1

i 2 b

(t2i 2 t2i2 1) (EQ 16)


E(TV) 5b

2b 1 1(EQ 17)

Therefore

as b ` E(TV)1

2 (EQ 18)



















P 5p00 p01

p10 p11(EQ 1)




p0 5 p0p00 1 p1p10

p1 5 p0p01 1 p1p11

p0 1 p1 5 1

(EQ 2)

and therefore

p0 5p10

p01 1 p10

p1 5p01

p01 1 p10

(EQ 3)









p00 5n 2 1 2 2b

n 2 1 2 b

p01 5b

n 2 1 2 b

(EQ 4)



p10 5 1 and p11 5 0 (EQ 5)


p0 5n 2 1 2 b

n 2 1(EQ 6)

and

p1 5b

n 2 1

where 2b n 2 1


p0 5b

n 2 1(EQ 7)

and

p1 5n 2 1 2 b

n 2 1



pL(b) 5b

n 2 1(EQ 8)

and

pH (b) 5n 2 1 2 b

n 2 1





presence (p)




b

n 2 1 p

1

2 (EQ 9)





P(low 2 presence)

5

n 2 1 2 2(b 2 k)

n 2 1 2 b 1 k

b 2 k

n 2 1 2 b 1 k

1 0(EQ 10)

with

pL(b) 5b 2 k

n 2 1(EQ 11)

and

P(high 2 presence)

5

k

b

b 2 k

b

b 2 k

n 2 1 2 b

n 2 1 2 2b 1 k

n 2 1 2 b

(EQ 12)

with

PH (b) 5n 2 1 2 b

n 2 1 (EQ 13)


P(low 2 presence)

5

n 2 1 2 2(b 2 s1 2 k)

n 2 1 2 (b 2 s1) 1 k

b 2 s1 2 k

n 2 1 2 (b 2 s1) 1 k

1 0

(EQ 14)

and

P (high 2 presence)

5

k

b 2 sn

b 2 sn 2 k

b 2 sn

b 2 s1 2 k

n 2 1 2 (b 2 sn)

n 2 1 1 k 2 2b 1 s1 1 snn 2 1 2 (b 2 sn)

(EQ 15)





4 Experiment








ObjectDimension(m)










Lowactivity

Highactivity






















piece












5 R esu lt s

5 1 General






















t 5 199(t16 5 2120 at 5)




High p 5 2 091 1 1107 HM(se 5 490)

226

Low p 5 087 1 034 HM(se 5 478)

007 (ns)















6 Discussion






High p 5 2 190 1 1911 HM(se 5 336)

569

Low p 5 087 1 034 HM(se 5 292)

011 (ns)


2131 at 5 Overall R2 5 073

























7 Conclusions




Regressionp 5

t-value


038 2 091 1 1107 HM 226


042 2 164 1 1700 HM 341


026 2 023 1 778 HM 146


077 2 146 1 1623 HM 701


043 2 164 1 1667 HM 320











Ack now ledgments



References








































TV 5 oi 5 1

i 2 b

(t2i 2 t2i2 1) (EQ 16)


E(TV) 5b

2b 1 1(EQ 17)

Therefore

as b ` E(TV)1

2 (EQ 18)








p00 5n 2 1 2 2b

n 2 1 2 b

p01 5b

n 2 1 2 b

(EQ 4)



p10 5 1 and p11 5 0 (EQ 5)


p0 5n 2 1 2 b

n 2 1(EQ 6)

and

p1 5b

n 2 1

where 2b n 2 1


p0 5b

n 2 1(EQ 7)

and

p1 5n 2 1 2 b

n 2 1



pL(b) 5b

n 2 1(EQ 8)

and

pH (b) 5n 2 1 2 b

n 2 1





presence (p)




b

n 2 1 p

1

2 (EQ 9)





P(low 2 presence)

5

n 2 1 2 2(b 2 k)

n 2 1 2 b 1 k

b 2 k

n 2 1 2 b 1 k

1 0(EQ 10)

with

pL(b) 5b 2 k

n 2 1(EQ 11)

and

P(high 2 presence)

5

k

b

b 2 k

b

b 2 k

n 2 1 2 b

n 2 1 2 2b 1 k

n 2 1 2 b

(EQ 12)

with

PH (b) 5n 2 1 2 b

n 2 1 (EQ 13)


P(low 2 presence)

5

n 2 1 2 2(b 2 s1 2 k)

n 2 1 2 (b 2 s1) 1 k

b 2 s1 2 k

n 2 1 2 (b 2 s1) 1 k

1 0

(EQ 14)

and

P (high 2 presence)

5

k

b 2 sn

b 2 sn 2 k

b 2 sn

b 2 s1 2 k

n 2 1 2 (b 2 sn)

n 2 1 1 k 2 2b 1 s1 1 snn 2 1 2 (b 2 sn)

(EQ 15)





4 Experiment








ObjectDimension(m)










Lowactivity

Highactivity






















piece












5 R esu lt s

5 1 General






















t 5 199(t16 5 2120 at 5)




High p 5 2 091 1 1107 HM(se 5 490)

226

Low p 5 087 1 034 HM(se 5 478)

007 (ns)















6 Discussion






High p 5 2 190 1 1911 HM(se 5 336)

569

Low p 5 087 1 034 HM(se 5 292)

011 (ns)


2131 at 5 Overall R2 5 073

























7 Conclusions




Regressionp 5

t-value


038 2 091 1 1107 HM 226


042 2 164 1 1700 HM 341


026 2 023 1 778 HM 146


077 2 146 1 1623 HM 701


043 2 164 1 1667 HM 320











Ack now ledgments



References








































TV 5 oi 5 1

i 2 b

(t2i 2 t2i2 1) (EQ 16)


E(TV) 5b

2b 1 1(EQ 17)

Therefore

as b ` E(TV)1

2 (EQ 18)





b

n 2 1 p

1

2 (EQ 9)





P(low 2 presence)

5

n 2 1 2 2(b 2 k)

n 2 1 2 b 1 k

b 2 k

n 2 1 2 b 1 k

1 0(EQ 10)

with

pL(b) 5b 2 k

n 2 1(EQ 11)

and

P(high 2 presence)

5

k

b

b 2 k

b

b 2 k

n 2 1 2 b

n 2 1 2 2b 1 k

n 2 1 2 b

(EQ 12)

with

PH (b) 5n 2 1 2 b

n 2 1 (EQ 13)


P(low 2 presence)

5

n 2 1 2 2(b 2 s1 2 k)

n 2 1 2 (b 2 s1) 1 k

b 2 s1 2 k

n 2 1 2 (b 2 s1) 1 k

1 0

(EQ 14)

and

P (high 2 presence)

5

k

b 2 sn

b 2 sn 2 k

b 2 sn

b 2 s1 2 k

n 2 1 2 (b 2 sn)

n 2 1 1 k 2 2b 1 s1 1 snn 2 1 2 (b 2 sn)

(EQ 15)





4 Experiment








ObjectDimension(m)










Lowactivity

Highactivity






















piece












5 R esu lt s

5 1 General






















t 5 199(t16 5 2120 at 5)




High p 5 2 091 1 1107 HM(se 5 490)

226

Low p 5 087 1 034 HM(se 5 478)

007 (ns)















6 Discussion






High p 5 2 190 1 1911 HM(se 5 336)

569

Low p 5 087 1 034 HM(se 5 292)

011 (ns)


2131 at 5 Overall R2 5 073

























7 Conclusions




Regressionp 5

t-value


038 2 091 1 1107 HM 226


042 2 164 1 1700 HM 341


026 2 023 1 778 HM 146


077 2 146 1 1623 HM 701


043 2 164 1 1667 HM 320











Ack now ledgments



References








































TV 5 oi 5 1

i 2 b

(t2i 2 t2i2 1) (EQ 16)


E(TV) 5b

2b 1 1(EQ 17)

Therefore

as b ` E(TV)1

2 (EQ 18)




4 Experiment








ObjectDimension(m)










Lowactivity

Highactivity






















piece












5 R esu lt s

5 1 General






















t 5 199(t16 5 2120 at 5)




High p 5 2 091 1 1107 HM(se 5 490)

226

Low p 5 087 1 034 HM(se 5 478)

007 (ns)















6 Discussion






High p 5 2 190 1 1911 HM(se 5 336)

569

Low p 5 087 1 034 HM(se 5 292)

011 (ns)


2131 at 5 Overall R2 5 073

























7 Conclusions




Regressionp 5

t-value


038 2 091 1 1107 HM 226


042 2 164 1 1700 HM 341


026 2 023 1 778 HM 146


077 2 146 1 1623 HM 701


043 2 164 1 1667 HM 320











Ack now ledgments



References








































TV 5 oi 5 1

i 2 b

(t2i 2 t2i2 1) (EQ 16)


E(TV) 5b

2b 1 1(EQ 17)

Therefore

as b ` E(TV)1

2 (EQ 18)










Lowactivity

Highactivity






















piece












5 R esu lt s

5 1 General






















t 5 199(t16 5 2120 at 5)




High p 5 2 091 1 1107 HM(se 5 490)

226

Low p 5 087 1 034 HM(se 5 478)

007 (ns)















6 Discussion






High p 5 2 190 1 1911 HM(se 5 336)

569

Low p 5 087 1 034 HM(se 5 292)

011 (ns)


2131 at 5 Overall R2 5 073

























7 Conclusions




Regressionp 5

t-value


038 2 091 1 1107 HM 226


042 2 164 1 1700 HM 341


026 2 023 1 778 HM 146


077 2 146 1 1623 HM 701


043 2 164 1 1667 HM 320











Ack now ledgments



References








































TV 5 oi 5 1

i 2 b

(t2i 2 t2i2 1) (EQ 16)


E(TV) 5b

2b 1 1(EQ 17)

Therefore

as b ` E(TV)1

2 (EQ 18)




















piece












5 R esu lt s

5 1 General






















t 5 199(t16 5 2120 at 5)




High p 5 2 091 1 1107 HM(se 5 490)

226

Low p 5 087 1 034 HM(se 5 478)

007 (ns)















6 Discussion






High p 5 2 190 1 1911 HM(se 5 336)

569

Low p 5 087 1 034 HM(se 5 292)

011 (ns)


2131 at 5 Overall R2 5 073

























7 Conclusions




Regressionp 5

t-value


038 2 091 1 1107 HM 226


042 2 164 1 1700 HM 341


026 2 023 1 778 HM 146


077 2 146 1 1623 HM 701


043 2 164 1 1667 HM 320











Ack now ledgments



References








































TV 5 oi 5 1

i 2 b

(t2i 2 t2i2 1) (EQ 16)


E(TV) 5b

2b 1 1(EQ 17)

Therefore

as b ` E(TV)1

2 (EQ 18)













5 R esu lt s

5 1 General






















t 5 199(t16 5 2120 at 5)




High p 5 2 091 1 1107 HM(se 5 490)

226

Low p 5 087 1 034 HM(se 5 478)

007 (ns)















6 Discussion






High p 5 2 190 1 1911 HM(se 5 336)

569

Low p 5 087 1 034 HM(se 5 292)

011 (ns)


2131 at 5 Overall R2 5 073

























7 Conclusions




Regressionp 5

t-value


038 2 091 1 1107 HM 226


042 2 164 1 1700 HM 341


026 2 023 1 778 HM 146


077 2 146 1 1623 HM 701


043 2 164 1 1667 HM 320











Ack now ledgments



References








































TV 5 oi 5 1

i 2 b

(t2i 2 t2i2 1) (EQ 16)


E(TV) 5b

2b 1 1(EQ 17)

Therefore

as b ` E(TV)1

2 (EQ 18)




















t 5 199(t16 5 2120 at 5)




High p 5 2 091 1 1107 HM(se 5 490)

226

Low p 5 087 1 034 HM(se 5 478)

007 (ns)















6 Discussion






High p 5 2 190 1 1911 HM(se 5 336)

569

Low p 5 087 1 034 HM(se 5 292)

011 (ns)


2131 at 5 Overall R2 5 073

























7 Conclusions




Regressionp 5

t-value


038 2 091 1 1107 HM 226


042 2 164 1 1700 HM 341


026 2 023 1 778 HM 146


077 2 146 1 1623 HM 701


043 2 164 1 1667 HM 320











Ack now ledgments



References








































TV 5 oi 5 1

i 2 b

(t2i 2 t2i2 1) (EQ 16)


E(TV) 5b

2b 1 1(EQ 17)

Therefore

as b ` E(TV)1

2 (EQ 18)















6 Discussion






High p 5 2 190 1 1911 HM(se 5 336)

569

Low p 5 087 1 034 HM(se 5 292)

011 (ns)


2131 at 5 Overall R2 5 073

























7 Conclusions




Regressionp 5

t-value


038 2 091 1 1107 HM 226


042 2 164 1 1700 HM 341


026 2 023 1 778 HM 146


077 2 146 1 1623 HM 701


043 2 164 1 1667 HM 320











Ack now ledgments



References








































TV 5 oi 5 1

i 2 b

(t2i 2 t2i2 1) (EQ 16)


E(TV) 5b

2b 1 1(EQ 17)

Therefore

as b ` E(TV)1

2 (EQ 18)


























7 Conclusions




Regressionp 5

t-value


038 2 091 1 1107 HM 226


042 2 164 1 1700 HM 341


026 2 023 1 778 HM 146


077 2 146 1 1623 HM 701


043 2 164 1 1667 HM 320











Ack now ledgments



References








































TV 5 oi 5 1

i 2 b

(t2i 2 t2i2 1) (EQ 16)


E(TV) 5b

2b 1 1(EQ 17)

Therefore

as b ` E(TV)1

2 (EQ 18)











Ack now ledgments



References








































TV 5 oi 5 1

i 2 b

(t2i 2 t2i2 1) (EQ 16)


E(TV) 5b

2b 1 1(EQ 17)

Therefore

as b ` E(TV)1

2 (EQ 18)



References








































TV 5 oi 5 1

i 2 b

(t2i 2 t2i2 1) (EQ 16)


E(TV) 5b

2b 1 1(EQ 17)

Therefore

as b ` E(TV)1

2 (EQ 18)












TV 5 oi 5 1

i 2 b

(t2i 2 t2i2 1) (EQ 16)


E(TV) 5b

2b 1 1(EQ 17)

Therefore

as b ` E(TV)1

2 (EQ 18)



A Virtual Presence Counter - Publications...

Documents

Transcript of A Virtual Presence Counter - Publications...