CALIFORNIA STATE UNIVERSITY, NORTHRIDGE SIMUI,TANEOUS PROCESSING OF AUDIO AND VISUAL INPUTS IN A VIGILANCE TASK A thesis submitted in partial satisfaction of the requirements for the degree of Master of Arts in Psychology Human Factors/Applied Experimental by Francine Harriet Landau August, 1981 The Thesis of Francine Harriet Landau is approved: Dr. vVilliami:Hlsoncjoft Dr. Tyler Bla}e /Dr. M~J>CSanders California State University, Northridge i TABLE OF CONTENTS Page ABSTRACT .. .. i i i It~TRODUCTION •••••••..•.•.••••.•.•••.••••.•• 1 METHOD ...... . I I I I I I I I I I I I I I I I I I I I I I I I I I I I I ............................... .7 Apparatus . .. . ... . . . . . .. . . . . . .. . . ... . .. . 7 Subjects. ................... .7 Materials. .. . . . . . . . . . . . .. .... . 8 Design .•.. Stimulus Procedure ........•.••.•....•....•..... 1 0 RESULTS •••••...•••••......•. I •••••• I I ••••• 11 Table 1 ................................ 13 Table 2 . • • • • • • • • • • • • Ill •••••••••••••••• .14 Figure 1 ............................... 15 F i gu.re 2 • • • . • • • • , • • . • . • • . • ., • • • . • . . . . . . 1 7 Figure J . Figure 4. DISCUSSION. • • • • • • • • • • • • • 0 •••••••••••••• .18 .. . ... . . .. . .. .. .. . ......... . .19 .............................. .20 REFERENCES • .. • • • • • • • • • • • • • • • • • .............. 3 2 ii ABSTRACT SIMULTANEOUS PROCESSING OF AUDIO AND VISUAL INPUTS IN A VIGILANCE TASK by Francine Harriet Landau Master of Arts in Psychology Percent of correct detections was used to investigate whether parallel processing was possible with auditory and visual stimuli being presented simultaneously. Sixty-four male and female college students participated in a vigilance task. Three factors, type of stimulus, signal pres- entation, and response were manipulated in a betweengroups factorial design. Stimuli were verbal or psycho- physical; responses were verbal or manual; and signals were presented synchronously (both input modes) or asynchronously (one input mode). It was found that semantic stimuli were harder for subjects to process accurately than psychophysical stimuli and that detection scores dropped in the second trial block, reflecting a vigilance decrement. Asynchronously presented auditory targets ir1 the second half of the session elicited the worst detection performance. Several phenomena such as habituation, visual dominance, and semantic interference were iii discussed which could account for the poor performance of that group. Evidence suggests that subjects were not processing information in a parallel fashion but instead were quickly switching attention between inputs or using other optional attentional strategies that were available as a result of the implementation of signal presentation. iv A classic question in the attention and performance literature is whether two simultaneous inputs can be processed at the same time. Broadbent's filter theory (1958) proposed that there is a central processing mechanism, i.e., attention, which is capable of handling information from only one "channel" at a time. To deal with simultaneous multiple inputs, attention is switched from channel to channel, sampling information for a brief, finite interval from each channel sequentially. Atten- tion is switched by a filter which handles inputs serially. That is, one input is processed first, and the filter only later retrieves the other item from storage. If processing of the first stimulus in the central system takes too long, the second stimulus will be lost from sensory or memory storage. In 1954 Broadbent in- vented the split-span experiment in which subjects monitored, recalled, shadowed (verbally repeated inputs), or replied to one or both of two simultaneous speech messages. Much of the research supporting Broadbent's theory was done with this kind of experiment (Ivlowbray, 1962, 1964; Moray & O'Brien, 1967; Trei.sman & Geffen, 1968). The predominant finding in these experiments was that subjects were only able to process information from one channel at a time or that information from dual channels was processed serially. In a typical split-span experiment, 'rreisman and 1 2 Riley (1969) had subjects shadow digits from one of two dichotically presented lists and to detect infrequent letters on either ear. The subjects were to i~mediately stop shadowing when they heard the critical item. Sub- jects detected 76 percent of letters on the attended ear and 33 percent on the other. Although this result sup- ported Broadbent's sequential processing hypothesis, researchers felt that the experiment demonstrated that attention cannot be divided between concurrent stimuli if the listener is biased toward one channel by the instruction to shadow one of the messages. A subsequent study by Tulving and Lindsay in 1970 corroborated this result. In contrast to these experiments, other studies demonstrated that parallel processing of concurrent verbal stimuli was possible. A study by Moray (1959) showed that people often noticed their spoken name while they were listening to another message. l.'Unio and Kahneman (1973) found that reaction time was approximately the same in both a divided attention and focused attention task. Further, subjects detected 77 percent of the targets in the divided attention task in which they had to press a key whenever a target word was heard while listening to dichotically presented word lists. Although evidence for parallel processing of diehotically presented stimuli is sparse, it is clear that under J p ' some conditions we can cope with two informative inputs at least partly in parallel. The variations among type of stimulus, response, and results for these experiments were so great that Treisman (1969) postulated a new hypothesis to help structure the heterogeneous findings. She sug- gested that independent analyzers (Sutherland, 1959) code different aspects or dimensions of incoming stimuli such as color, pitch, loudness, and spatial location. If tasks involve inputs which converge on the same analyzing system, it would seem that serial processing must take place. On the other hand, if separate analyzing systems are invoked by one or more inputs, parallel processing may be effected. This theory predicts that the subject should be able to monitor two messages as easily as one, provided that the messages do not reach the same analyzer. Yntema and Schulman (1970) supported this logic when they found that subjects had less difficulty keeping track of states of different variables than monitoring different states of the same type of variable. Using this hypothesis as a departure, experiments began to include the case in which different modalities were used as input sources for the divided attention taslcs. Allport, Antonis, and Reynolds (1972) found that subjects could carry out two tasks in parallel (shadowing speech while sight reading music) with little or no interference. Although it is evident that parallel 4 processing can take place, researchers were interested in determining when dual inputs were most efficient, and for what kind of task. Although the findings in experiments using complex verbal stimuli were inconsistent, there was a preponderance of evidence supporting the superiority of dual mode inputs in vigilance tasks when stimuli were simplistic and psychophysical, e.g., audio stimuli that changed frequency or intensity, visual stimuli that became brighter or changed color. Typically, these experiments had sub- jects monitoring inputs for extended periods (90 minutes) and the subjects were trained for the task to eliminate As early as 1958, Klemmer demon- any practice effects. strated that a simultaneous presentation of auditory and visual signals (each reinforcing the other) against a background noise condition produced more efficient reaction time and higher detection rates than either a single visual or audio signal. \j) and~olquhoun Tyler, Waag, and Holcomb (1972) (1975) all found that redundant inputs from both audio and visual sources elicited superior detection performance in a vigilance task.~ In all these experiments the dual mode condition was represented by simultaneous redundant presentation of stimuli and signals on both audio and visual inputs, i.e., the signal was presented synchronously on both input sources every time. In a 1965 study, McGrath used a 5 different implementation of signal presentation: back- ground stimuli were presented in both modes once every three seconds but the signal was presented on only one source aperiodically (asynchronously). His investigation also included a distinction between "easy" detection tasks in which subjects detected the signal 90 percent of the time versus a "difficult" detection task in which subjects could detect the signal only 65 percent of the time. With stimuli being presented simultaneously and signals asynchronously, McGrath found vigilance performance enhanced for the easier task only. The lack of signal re- dundancy and the uncertainty about where the signal would originate may have introduced additional processing demands which would have been responsible for the differential enhancement of detection. In a 1979 dual mode study, Whitaker investigated how response time in a monitoring task was influenced by set size of response choices. The implementation of the dual task condition included both a verbal response to an audio input and the addition of another visual tracking task which was done concurrently. This evidently exceeded the capacity of the subject to perform parallel processing because response time was much slower for the dual task condition. Thus, the range of parameters facilitating parallel processing in vigilance tasks may be Yery small. The enhancement effect of dual inputs found in the 6 vigilance literature is in direct opposition to results with verbal split-span experiments where reliable parallel processing of dichotically presented inputs has not been demonstrated. One important difference seems to be whether signals appeared simultaneously (synchronously) or alternately (asynchronously) on both input sources. Fbr this reason the present study investigated both synchronous and asynchronous signals presented in a dual mode task. It was hypothesized that synchronous presen- tation of signals would enhance detection performance. Another aspect of dual mode tasks which appeared to influence the possibility of efficient parallel processing was the type of stimulus used. In the verbal learning literature previously reviewed, the content of research was predominantly semantic. stimuli were psychophysical. In the vigilance literature, These two stimuli types were combined in the present study as two levels of a stimulus factor. It was hypothesized that simplistic inputs would be easier to process than semantic inputs. Each area of research was also characterized by one set of responses. The verbal learning literature typically used variations of verbal responses while, in the vigilance literature, motor responses were used most often. Using the empirical evidence from the vigilance experiments as a rationale, it was predicted that motor responses would interfere less than verbal responses with simultaneous processing. Method Subjects The subjects were 19 male and 45 female college sophomores and juniors with normal vision and hearing who were required to accumulate experimental credits for their Psychology courses. Subjects were screened on a pretest of the task and had to obtain a detection score of 60 percent in order to be accepted into the experiment. Only one person was dropped for this reason. Apparatus The equipment included a Panasonic TV monitor and p;ortable video recorder NV3085 with tape meter; two pairs of Telephics TDH-49 monaural earphones with separate volume controls, one for the subject and one for the experimenter; a two button response box connected to a 12.7 em x 17.8 em lightboard indicator with a red and green light bulb. Each push button, labeled with the appropriate res- ponse word ("see" or "hear") was connected to a light bulb which would light when the subject pushed the button. power supplies were used: Two a Panasonic AC adaptor NV-B40 and a BRS electronics model PS-12b. Design Three independent variables, type of stimulus, signal presentation, and response were investigated in a 2x2x2 between groups factorial design. 7 Since stimuli were 8 presented both visually and aurally, the two levels of stimulus type were dots and printed words for visual presentation, and tones and spoken words for auditory presentation. The two levels of signal presentation were syn- chronous and asynchronous; in the synchronous condition, two signals were presented simultaneously, one visually and one auditorily. In the asynchronous condition, the signals were presented aperiodically, either on the TV or over the earphones, but never at the same time. Responses could be either button pushes or verbal reports ("see .. or "hear"). The assignment of subjects to conditions and the order of words and signals was randomly determined. Sig- nals were sequenced during the session so that no temporal or spatial pattern was discernible and they occurred, on the average, once every 2.15 minutes. The dependent var- iable was percent of correct detections. Stimulus Materials Four tapes, each .30 minutes long, were produced for combinations of the two independent variables, stimulus type and signal presentation type. Each tape contained 900 stimulus presentations for both the visual and auditory modes. t Each stimulus was presented for a duration of second and the interstimulus interval was 1.5 seconds. The following is a description of the contents of the four tapes. Synchronous-Dot-Tone. A 12.7 mm black dot subtending 9 a visual angle of 1° 12' appeared on the center of a white background on the TV screen every two seconds. At the same time, a 300 Hz, 50 db. tone was played every two seconds over the earphones. The signals in this condition were a 19.05 mm black dot, subtending a visual angle of 1° 48' and a 300 Hz, 70 db. tone. Fourteen signals were aural and fourteen were visual, totaling 28. These sig- nals were always presented at the same time. Asynchronous-Dot-Tone. A 12.7 mm black dot appeared on a white background every two seconds on the TV screen while a 300 Hz, 50 db. tone was heard through the earphones at the same time. A 19.05 mm black dot, the visual signal, appeared seven times during the 30 minute run. A 300Hz, 70 db. tone, the audio signal, occurred seven times during the session but never at the same time as the visual signal. Synchronous-Verbal-Verbal. A monosyllabic word, e.g., "know", "dish", "cool", appeared every two seconds in the center of the TV screen. At the same time a monosyllabic word was heard over the earphones at exactly the same rate and duration. The signal was a word of the animal class such as "dog", "snake", or"cat". The words subtended an average visual angle of 4° 10'. There were 14 visual and 14 audio signals always presented at the same time, totaling 28. As~nchronous-Verbal-Verbal. The stimuli and signals 10 were the same as the previous condition, however, a total of 14 signals were presented, seven by earphones, and seven on the TV. Audio and visual signals never appeared at the same time. Illumination in the room while subjects were taking part in the experiment was 2766.3 lx; the luminance contrast between stimuli and background in the dot-tone and verbal conditions were 94 and 90 percent respectively. Because of an afterimage problem with the visual stimuli, a second light field of the same intensity as the background came on at the offset of the stimulus and went off at the onset of the stimulus. Because of an anticipated summation phenomenon on the auditory tapes, a 25 db., 1kHz bandwidth of white noise was played during the ISI. This white noise was also present on the verbal stimulus tapes. Procedure The subjects were asked to sit directly in front of the TV screen which was approximately 69 em away from their face. The type of visual stimulus was displayed on a card, either a dot or word, and then the type of signal was explained. For the dot-tone condition, the signal was the larger dot; for the word condition, it was an animal word. The appropriate response upon iden- tifying a signal was explained and then subjects practiced identifying four visual signals. After practice with the visual stimuli, the auditory stimuli and signals were demonstrated using the earphones. viTi th the TV off, the subjects practiced identifying four auditory signals, either a louder tone, or an animal word. If they had trouble hearing and/or seeing two of four signals, they were thanked, given credit, and dropped from the experiment. After the subjects practiced monitoring each mode separately, they practiced monitoring both modes together for seven minutes. If they failed to detect at least 60 percent of the targets, they were not used in the experiment. The experimenter then changed tapes while the par- ticipants took a five minute break in another room. The experimental run took 30 minutes, and at the end of the session, subjects were given a debrief sheet. Results Using a three-way factorial ANOVA, it was found, as hypothesized, that dot-tone stimuli were easier to process than verbal stimuli, F (1,56) = 205.323, p <.001. Since 19 of the 32 subjects in this condition scored 100 percent, it was felt that the homogeneity of variance assumption for ANOVA could not be met. In addition, this ceil- ing effect would have made all other statistical results suspect. Therefore, only the verbal stimulus data were used and a new four-way ANOVA was performed. The two re- maining between subject factors--response mode and signal presentation--were combined with two repeated measures 11 12 --trial blocks and stimulus mode. For the trial blocks factor, the experimental session was divided in half, giving two detection scores. These detection scores were again separated into auditory and visual target detection scores for the stimulus mode factor. Although false alarm data was collected, the highest group mean for false alarms was 7.0 in a session that contained 900 stimulus presentations. This data was not included in any further analysis since these responses were too infrequent to yield meaningful interpretations. Using percent of cor- rect detections as the dependent variable, Table 1 contains the means and standard deviations for the four-way analysis data. Table 2 contains the F ratios ~s well as eta 2 for the four-way ANOVA data. A significant trial blocks by response interaction was found which is illustrated in Figure 1. A Newman- Keuls test indicated that correct verbal responses dropped significantly from trial block 1 to trial block 2, Grit. Diff. (32) = 17.19, p<. .05, while correct button responses did not change with time on task. All other comparisons among trial block and response groups were not significant. Although the next two significant interactions identified by the ANOVA will be reported here, it should be noted that they are in turn modified by another variable 13 Table 1 Mean Percent and Standard Deviations for Correctly Identified Audio and Visual Targets in Two Trial Blocks Trial Block 2 1 Synchronous Stimulus :.:ode Audio 7isual Response Mode Verbal 92.85 ( Button 87.49 Verbal Button 7.64)a 73.2 (19.37) (14.17) 91.08 (20.10) 92.85 ( 7.64) '?4.99 (19.84) 78.56 (22.90) 87.49 (14.17) Asynchronous Audio Tisual 1 Verbal 9.5.84 {11. 77) 58.J4 (34. 51) Button 8J.J4 (1.5.42) 62.49 (JJ,05) Verbal 84.J8 (22.90) 90.63 (12.94) Button 81.78 (18.35) 79.69 (18.32) aStandard deviations in parentheses. 14 Table 2 ANOVA Ratios for Four-Way Target Detection Study Source s~ eta2 .!U: !§. ! 62.)0 1 62.)0 .09 Presentation {Syn-Asyn) 88:3.58 1 88).58 1.22 1. 5:3 Response x Presentation 5:31.79 1 531.79 ,7J .92 20261.4.8 28 72:3.62 Trials (Blk 1-Blk .2) :3135.:33 1 :31:35-JJ 11.77 .002 5.4-J Trials x Response 1701.6:3 1 1701.63 6.)9 .017 2.95 Trials x Presentation 425.96 1 425.96 1.60 .74 Trials x Response x Presentation 869.97 1 869.97 ).27 1.50 7456.97 28 266.)2 Mode (Audio-Visual) :3:31.21 1 :3:31.21 1.21 .57 Mode x Response 189.88 1 189.88 .69 ,JJ l11J.JJ 1 111J.JJ 4,06 41.52 1 41.52 .15 Error 768?.52 28 274.55 Trial x Mode 2424.69 1 2424.69 229.78 1 229.78 15JJ.89 1 15JJ.89' 5.11 4-<>7.20 1 407.20 1.36 Error 8404.19 28 )00.15 Total 57692.22 127 Response · (Manual-Verbal) Error Error Mode x Presentation Mode x Response x Presentation Trial x Mode x Response Trials x Mode x Presentation Trial x Mode x Response x Presentation l2. .1 .054 1.93 .07 a.oa .ooa 4,2 .40 .77 .0)2 2.66 .70 ~ 15 95 93 91 91.48 89 87 Percent 85 Correct 83 Detections 81 82.79 80.23 79 77 75 74.29 73 Trial Block Trial Block 1 2 Figure 1. Interaction of trial blocks and response type with group means. 16 in a significant three-way interaction. Stimulus mode and signal presentation interact significantly and are graphed in Figure 2. Synchronous audio and visual signals were equally hard to detect as were asynchronous visual signals. However, the percent of correctly detected asyn- chronous audio signals was much lower than any of the other three groups. Since a third variable, trial blocks, further modifies this interaction between stimulus mode and signal presentation, no other ad hoc comparison tests were made. A significant trial blocks by stimulus mode interaction was also found and this is illustrated in Figure J. Audio signals were harder to detect in trial block 2 than in trial block 1. Subjects were able to detect visual signals equally well in trial block 1 or trial block 2. Again, the probability of detecting the two types of stimuli depended on whether the subject was identifying asynchronous or synchronous signals in trial block 1 vs. trial block 2. in Figure 4. This three-way interaction is illustrated The interaction of trial blocks and stimulus mode was not significant with synchronously presented signals but was with asynchronously presented signals. A Newman-Keuls test indicated that asynchronous audio signals became much harder to correctly identify in trial block 2 than asynchronous visual signals in that trial block, Crit. Diff. (16) = 24.75, J2.C..05. Both synchronous 1'1 95 93 91 89 87 Percent 85 Correct 83 Detections 86.15 Visual 83.47 84.12 81 79 77 75 75.0 73 Asynchronous Synchronous Presentation Figure 2. Interaction of signal presentation and stimulus mode with group means. 18 91 89 89.88 87 85 Percent 83 Correct 81 Detections 79 Visual 84.39 83.20 77 75 73 Audio 71 71.26 69 Trial Block Trial Block 1 2 Figure 3. Interaction of trial blocks and stimulus mode with group means. 19 91 89 87 ·as 90.17 85.7 a:; Percent Trial Block 2 Trial Block 1 a. Synchronous Presentation Correct Detections 82.14 81.18 81 95 9:3 91 89 87 89.59 as a:; 81 79 77 75 73 71 -Visual 85.16 8J.07 69 67 65 63 61 59 60.41 Trial Block 2 Trial Block 1 b. Asynchronous Presentation Figure 4. Intaraction of trial blocks, signal presentation, and stimulus mode with group means. audio and visual signals were slightly harder to detect in trial block 1 than trial block 2, but not significantly so. The greatest drop in performance occurred between trial block 1 and trial block 2 with asynchronous audio targets, Crit. Diff. (16) = 29.18, p ~ .05. Performance with asynchronous visual targets did not change significantly from trial block 1 to trial block 2. Thus, the hardest signals to detect in this experiment were asynchronous audio signals in the second trial block. The three variables, stimulus mode, signal presentation, and trial blocks, modified the effect of each other so that differential performance occurred depending on the specific combination of levels experienced. Discussion The original hypothesis that semantic material is more difficult to process than simple stimuli such as dots and tones was supported in this experiment. The mean per- cent of correct detections for verbal stimuli was 82.1 as opposed to 95.8 for the dot-tone stimuli. As previously mentioned, only verbal data were used in the analysis because of a ceiling effect in the dot-tone data. The hypo- thesis that motor responses would have a generally positive influence on detection behavior was not supported in this experiment. In addition, no main effect of signal presentation was found. In general, the subjects in all groups, except one, had little difficulty processing 20 21 inputs from dual sources. The mean percent correct detec- tions ranged from 73.2 to 95.0. If subjects could not process both inputs, one would expect that performance for at least one input would be reduced to chance. This was not the case and these results support other research (Allport, Antonis, & Reynolds, 1972; Colquhoun, 1975; McGrath, 1965; McLeod, 1977; Treisman & Davies, 1973; Tyler, Waag, & Holcomb, 1972) which indicated that detection performance in dual mode tasks was as good or better than in single mode tasks. A significant effect of trial blocks indicated that performance in the first helf of the session was superior to performance in the second half. tions dropped from 92.1 to 86.0. Mean percent of detecEvidently, subjects can process simultaneous stimuli, but it is not clear for how long. This drop in accurate detection has been inter- preted as reflecting the commonly found vigilance decrement. This effect has been noted and explored since Mackworth first investigated the vigilance task in 1950. Many theories have been advanced to explain this phenomenon; all of them predict vigilance behavior fairly well but none of them explain all vigilance data (Loeb & Alluisi, 1980). A very cursory review of the current major theories explaining the vigilance decrement will be described here with no attempt to ascribe priority. Arousal and habituation are two neurological models 22 which have been investigated in the literature. Arousal is a measure of how wide awake the organism is, of how ready it is to react (Berlyne, 1960). The ascending retic- ular activating system and the diffuse thalamic projection system are portions of the brain thought to mediate arousal. There is an optimum level of arousal for per- formance, but with time on task, arousal tends to decline, with subsequent deterioration in performance. Habituation is a decline in perceptual sensitivity of the observer because of repeated exposure to non-signal stimuli (Kahneman, 1973), again producing a decrement in detection performance. In this context, habituation of the orienting response, a response to novel stimuli such as a signal, has been noted by researchers (Coquery, 1978; Kahneman, 1973; Mack, 1970; Stroh, 1971). As an indica- tion of neurological involvement, the magnitude of evoked potentials within the brain has been measured. Haider (1967) found a decline in amplitude of evoked cortical potentials with a concomitant decline in percent of correct detections. Other researchers have found similar results (Davies & Krkovic, 1965; Mackworth, 1970; Picton, Campbell, Baribeau-Braun, & Prouix, 1978; Stroh, 1971; Wilkinson, Morlock, & Williams, 1966). A third hypothesis suggests that subjects expect unrealistically high rates of signal presentations and that decrements occur because signal rates are lower 23 (Baker, 1959; Colquhoun & Baddeley, 1964; Craig, 1978: Deese, 1955; Thomas, 1971). A general probability match- ing trend of the subject becomes more obvious during the experimental run as he progressively tries to match his propensity to detect a stimulus with the ratio of signals to non-signals (Loeb & Alluisi, 1980). Using a signal detection framework for interpretation of vigilance decrements, Swets in 1977 concluded that subjects adopt a stricter criterion as sessions go on in which they learn to better identify non-signals from signals and are less inclined to detect a stimulus as a signal. Again, there is much support for this model (Broadbent & Gregory, 1965; Colquhoun, 1967a; Egan, Greenberg, & Schulman, 1961; Loeb & Alluisi, 1980). Most probably all these models reflect ongoing processes within the vigilance situation which could account for the drop in accuracy in this experiment. However, an investigation by Parasuraman (1979) seems particularly relevant to this study. He used similar parameters of task, load, and stimuli type to determine whether the, vigilance decrement was the result of a loss of perceptual sensitivity or a change in response criterion. Event rates of non-target stimuli were presented either once every two seconds or once every four seconds. The tasks were either successive discrimination, i.e., the signal was presented alone after many repetitions of non-signal stimuli, or simultaneous discrimination where the signal was presented within a field of non-signal stimuli. Parasuraman found that the decrement was due to a loss of perceptual sensitivity when signal discrimination loads memory (successive discrimination) and when stimulus events occur rapidly (once every two seconds). More interesting is the finding that with time on task, verbal responses had a negative influence on performance as opposed to motor responses which elicited similar performance for both time blocks. This result partially supports the initial hypothesis that verbal responses may be interferred with by verbal encoding of stimulus inputs. McLeod (1977) analyzed response interference in dual mode tasks and concluded that the similarity of task and response could predict the level of interference which would be reflected in poor performance. Using two levels of difficulty in an arithmetic task, he found no debilitating effect on manual responses while significant interference was obtained using verbal responses. The finding that dimensions of tasks that use common analyzers generate the most interference has been found by many investigators (Kinsbourne & Hicks, 1978; McLeod, 1977; Treisman & Riley, 1969). Evidently, as interference from verbal inputs and fatigue increased, verbal responses may have demanded more processing time which subjects did not have with the fast presentation of stimuli in this study. 25 The question of whether subjects can receive and process parallel inputs has to be addressed in determining the cause of the three~way interaction between trial blocks, stimulus mode, and signal presentation. Most in- vestigators support the notion that subjects can receive information in parallel but disagree on the possibility of higher order parallel processing. The evidence, here, suggests subjects can process dual inputs in the synchronous condition but their performance may have been due to optional attentional strategies and not to valid higher order processing in parallel. Because in the synchronous mode signals were always presented on both inputs at the same time, subjects became aware that when an audio signal was presented, a visual signal was also being presented. This cueing not only heightened short term attention but alleviated the demand for constant attention to both modes throughout the session. Subjects could switch attention from one mode to another at will, thereby attenuating the effects of habituation, boredom, and fatigue. This strat- egy may offer an explanation for the positive findings in the vigilance literature where redundant signals were efficiently processed in dual mode tasks. It may also explain why performance was maintained as well as it was in trial block 2 for those subjects seeing only synchronous signals. Subjects had much more difficulty processing 26 asynchronously presented audio signals in the second half of the experiment. It is more certain that subjects were forced to either process inputs in parallel in the asynchronous condition or switch attention very quickly (250 msec) between inputs because of the lack of any other effective strategy. In trial block 1 subjects were able to process both input sources but in trial block 2 performance seems to break down, especially for audio signals. There are two general phenomena which debilitate performance in vigilance tasks and both were present in the asynchronous condition. uncertainty. These are temporal and spatial Temporal uncertainty, the irregular pres- entation of infrequent signals, has been the subject of many vigilance experiments and has elicited fairly consistent results with both reaction time and correct percent of detections as dependent variables. When there is no pattern for signal presentation, subjects don't know how to allocate attention and this results in missed signals (Adams & Boulter, 1964; Alegria & Bertelson, 1970; Baker, 1959b; McCormack & Prysiazniuk, 1961; Webber & Adams, 1964). Although temporal uncertainty may have a transient arousal effect, it is debilitating in tasks of long duration. The concept of spatial uncertainty, not knowing on which source the signal will appear, has been investigated by Adams and Boulter (1964) in conjunction with the 27 effects of temporal uncertainty. Three signal sources were manipulated with differing levels of temporal certainty. It was found that latency was minimal for either spatial and temporal certainty, but maximal for both kinds of uncertainty. Kulp and Alluisi (1967) investigated the same phenomena and had concluded that the contributions of different sources of uncertainty were linearly additive. These two sources of uncertainty may explain why subjects were under cumulative stress by the second time block, but it does not explain why this stress interferred with the detection of audio signals only. This decrement in detection for auditory signals in vigilance tasks has been documented throughout the literature (Baker & Harabedian, 1962; Buckner & McGrath, 1965; Gruber, 1964; Rollins, Jr. & Hendricks, 1980). Parsuraman (1979) com- pared 13 dual input investigations finding a vigilance decrement. the s~~dies. He found an auditory decrement in eight of Craig, Colquhoun, and Corcoran (1976), attempting to account for this effect, proposed that, as sessions went on, subjects adopted a more conservative criterion for detecting auditory targets, with a concomitant drop in correct detection. Therefore, the worst performance in this experiment may have been the result of subjects using a stricter criterion for identifying audio signals at the same time signal presentation uncertainties were creating additional stress for the subjects. 28 Although these are plausible explanations for the poor processing of asynchronous audio signals, another hypothesis will be discussed to account for the drop in accuracy for aural signals. An assumption made by this researcher and supported by the slightly better detection performance of asynchronous visual signals in trial block 2 was that subjects consciously attended the visual display and processed visual inputs first. Posner, Nissen, and Klein (1976) have reported on the tendency of subjects to attend to visual inputs in preference to auditory inputs when both are presented. This appears to be related to the superior alerting capacity of the auditory system and the relatively weak capacity of the visual system to alert the organism. Subjects, assuming they will be auto- matically alerted by auditory inputs and compensating for the weak alerting capacity of visual inputs, deliberately keep their attention tuned to the visual inputs. In a 1974 study, Colvita asked subjects to press one key whenever a light came on and another key whenever a tone occurred. When the signals were presented simultaneously, the subjects generally responded to the visual inputs and were often unaware that an auditory signal had occurred. When informed that signals could occur on both modes, subjects still attended to the visual inputs and detection for those stimuli was better. If subjects were watching for the visual signal, they had approximately 250 msec 29 to receive and encode that stimulus. Within the next 250 msec the subject had to receive and encode the auditory stimulus. If this sequence was somehow disrupted, the visual signal had the higher probability of being processed properly. Interference in cognitive processing may have occurred at two levels, short-term memory and/or semantic analysis, and could have caused information from the second stimulus to be lost. Lindsay (1970) reported on a short-term memory study using dual inputs in which subjects were asked to report judgments about psychophysical stimuli in a specific order. The dimension reported second suffered a substantially larger impairment than the first dimension reported on a trial. In another dual task study, Lindsay (1970) commented that subjects perform more efficiently when attending to different dimensions of the same type of stimulus than attending to the same dimension of different stimuli. This implies that specific ana- lyzers with limited capacity are available, subsequently restricting the amount of processing per unit time. In this study, semantic analysis of the visual word, which was received first, may have interferred with the semantic analysis of the auditory word which was processed second. Therefore, problems with both short-term memory and semantic analyzers may have caused the decline in auditory detection performance. This result lends support to the single channel filter theory of Broadbent which states 30 that multiple inputs are serially processed. In summary, this experiment provides results which are congruent with both vigilance and verbal learning research utilizing dual inputs. With redundant signals, processing of dual inputs is possible because of alternative attentional strategies which optimize switching techniques. Information is not being processed simulta- neously from input to output. Motor responses typically used in vigilance experiments did not interfere with cognitive processing in that research and did not interfere with semantic processing in this study. When subjects in the asynchronous condition were forced to process inputs in parallel, they could do it for a short period of time. Again, it is hypothesized that they accomplished this by quickly switching attention. Interference in limited capacity analyzers or processors soon made this technique unsuccessful. The verbal learning experiments reflected this situation where both stimuli and responses were analyzed by the same cognitive structures and performance was subsequently degraded. Further research should attempt to isolate where in the input-output process semantic interference is most likely to occur, and whether coupling manual responses with semantic inputs would eliminate that interference and facilitate performance. In addition, evidence from this study implicates the auditory system as more vulnerable 31 to habituation with extended task time. This result should be explored and verified so that future investigations can account for this phenomenon when designing dual mode tasks. 32 References Adams, J.A., & Boulter, L.R. Spatial and temporal uncertainty as determinants of vigilance behavior. Journal of Experimental Psychology, 1964, £1, 127-131. Alegria, J., & Bertelson, P. Time uncertainty, number of alternatives and particular signal-response pair as determinants of choice reaction time. Acta Psychologica, 1970, ]1, 42-50. Allport, D.A., Antonis, B., & Reynolds, P. On the division of attention: A disproof of the single channel hypothesis. Quarterly Journal of ExDerimental Psychology, 1972, 24, 225-235. Baker, C.H. Attention to visual displays during a vlgllance task. British Journal of Psychology, 1959, 2Q, 30-36. (Acta Psychologica, 1970, 11• 122-131.) Baker, C.H. Towards a theory of vigilance. Canadian Journal of Psychology, 1959b, 1lr 35-42. Cited by D.R. Davies & G.S. Tune, Human Vigilance Performance, New York1 American Elsevier Publishing Company, Inc., 1969. Baker, C.H., & Harabedian, A. Performance in~ auditory~ vigilance task while simultaneousl* tracking a visual target. Technical Report No. 740- 2, 1962. Cited by J. Mackworth, Vigilance and Attention: A Signal Detection Anproach, Baltimore: Penguin Books, 1970. Berlyne, D.E. Conflict, arousal and curiosity. McGraw Hill, 1960. New York: Bernstein, I.H. Can we see and hear at the same time? Acta Psychologica, 1970, Jl, 21-35. Broadbent, D.E. Perception and communication. Pergamon Press, Ltd., 1958. London: Broadbent, D.E., & Gregory, M. Effects of noise and of signal rate upon vigilance analysed by means of decision theory. Human Factors, 1965, L• 155-162. Colquhoun, W.P., & Baddeley, A.D. The role of pretest expectancies in vigilance decrement. Journal of Experimental Psychology, 1964, 68, 156-160. JJ Colquhoun, W.P. Sonar target detection as a decision process. Journal of Applied Psychology, 1967a, ji, 187-190. Colquhoun, W.P. Auditory, visual, and dual mode displays~ for prolonged sonar monitoring in repeated sessions. Human Factors, 1975, !1(5), 425-4]7. Colvita, F.B. Human sensory dominance. Psychophysics, 1974, 16, 409-412. Perception and Coquery, J.M. Selective attention as a motor program. In J. Requin (Ed.), Attention and performance VI. Hillsdale, New Jersey: Lawrence Erlbaum Assoc., 1978. Craig, A., Colquhoun, W.P., & Corcoran, D.W. Combining evidence presented simultaneously to the eye and the ear: A comparison of some predictive models. Perception and Psychophysics, 1976, 12(6), 47]-484. Craig, A. Is the vigilance decrement simply a response adjustment toward probability matching? Human Factors, 1978, 20, 442-446. Davies, D.R., & Krkovic, A. Skin conductance, alpha activity and vigilance. American Journal of Psychology, 1965, ~. 304-306. Cited by D.R. Davies & G.S. Tune, Human Vigilance Performance, New York: American Elsevier Publishing Company, Inc., 1969. Davies, D.R., & Tune, G.S. Human vigilance Derformance. New York: American Elsevier Publishing Company, Inc., 1969. Deese, J. Some problems in the theory of vigilance. Psychological Review, 1955, 62, 359-386. Cited by J. Mackworth, Vigilance and Attention: A Signal Detection Approach, Baltimore: Penguin Books, 1970. Deutsch, J.A., & Deutsch, D. Attention: some theoretical considerations. Psychological Review, 1963, 1Q, 8090. Cited by D. Kahneman, Attention and Effort, Englewood Cliffs, New Jersey: Prentice-Hall, Inc., 1973. Egan, J.P., Greenberg, G.Z., & Schulman, A.I. Operating characteristics and the method of free response. Journal of the Acoustical Society of America, 1961, Jl, 993-1007. 34 Gruber, A. Sensory alternation and performance in a vigilance task. Human Factors, 1964, 6, 3-12. Haider, M. Vigilance, attention, expectation, and cortical evoked potentials. In A.F. Sanders (Ed.), Attention~ ~erformance, Amsterdam: North Holland Publishers, 19 7. Kahneman, D. Attention and effort. Hall, Inc., 1973. - New Jersey: Prentice- Keele, S. Attention and human performance. Pacific Palisades, California: Goodyear Publishing Inc., 1973. Klemmer, E.T. Timesharing between frequency-coded auditory and visual channels. Journal of Experimental Psychology, 1958, jQ, 229-235. Kulp, R.A., & Alluisi, E.A. Effects of stimulus-response uncertainty on watchkeeping performance and choice reactions. Perception and Psychophysics, 1967, ~, 511-515. LaBerge, D. Identification of two components of the time to switch attention: A test of a serial and a parallel mod~l of attention. InS. Kornblum (Ed.), Attention and performance IV, New York: Academic Press, 1973. Lindsay, P·.H. Multichannel processing in perception. In D. Mostofsky (Ed.), Attention: contemporary theory and analysis, New York: Appleton-Century-Crofts, 1970. . Loeb, M. On the analysis and interpretation of vigilance: some remarks on two recent articles by Craig. Human Factors, 1978, 20(4), 447-451. Loeb, M., & Alluisi, E.A. Theories of vigilance: A modern perspective. Proceedings of the 24th Annual Meeting of the Human Factors Society, 1980, 600-604. Mackworth, N.H. Researchs in the measurement of human performance. In L. Sinaiko (Ed.), Selected Papers gn Human Factors in the Design ~ Use of Control Systems, London: Dover Publications, 1955. Mackworth, J.F. Vigilance and attention: _A signal detection approach. Baltimore: Penguin Books, 1970. 35 McCormack, P.D., & Prysiazniuk, A.W. Reaction time and regularity of interstimulus interval. Perceptual and Motor Skills, 1961, 12• 15-18. McGrath, J.J. Performance sharing in an audio-visual vigilance task. Human Factors, 1965, 141-153. z, McLeod, P. A dual task response modality effect: support for multiprocessor models of attention. Quarterly Journal of Experimental Psychology, 1977, ~~ b51-667. Moray, N. Attention in dichotic listening' affective cues and the influence of instructions. Quarterl~ Journal of Experimental Psychologr, 1959, 11, 56- o. Moray, N., & O'Brien, T. Signal-detection theory applied to selective listening. Journal of-the Acoustical Society of America, 1967, 42, 765-772. Moray, N. Where is capacity limited? In A.F. Sanders (Ed.), Attention and performance 1, Amsterdam: Humanities Press, 1970. Mowbray, G.H. Simultaneous v1s1on and audition: the comprehension of prose passages with varying levels of difficulty. Journal of Experimental Psychology, 1953, 46, 365-372. Mowbray, G.H. Some human perceptual imits, APL Technical Digest, 1962, 21-24. Cited by D. Kahneman, Attention and Effort, Englewood Cliffs, New Jersey: PrenticeHall, Inc., 1973. Mowbray, G.H. Perception and retention of verbal information presented during auditory shadowing. Journal of the Acoustical Socie~ of America, 1964, 36, 1459- 1465. Ninio, A., & Kahneman, D. Reaction time in focused and in divided attention. Journal of Experimental Psychology, 1973 in press. Cited by D. Kahneman, Attention and Effort, Englewood Cliffs, New Jersey: Prentice-Hall, Inc., 1973. Parasuraman, R. Memory load and event rate control sensitivity decrements in sustained attention. Science, 1979, 205, 924-926. Picton, T.W., Campbell, K.B., Baribeau-Braun, F., & Prouix, M. The neurophysiology of human attention: A tutorial review. In J. Requin (Ed.), Attention and 36 performance IV, Hillsdale, New Jersey: Erlbaum Assoc., 1978. Lawrence Posner, M.I., Nissen, M.J., & Klein, R.M. Visual dominance: an information processing account of its origins and significance. Psychological Review, 1976, ~. 157-171. Proctor, R.W., & Proctor, J.D. Task modality and attention. Journal of Experimental Psychologt: Human Perception and Performance, 1979, 2{4), 10-624. Rollins, Jr., H.A., & Hendricks, R. Processing of words presented simultaneously to eye and ear. Journal of Experimental Psycholof): Human Perception and Performance, 1980, 6(1 , 99-109. Stroh, C.M. Viglance: the problem of sustained attention. New York: Pergamon Press, 1971. Sutherland, N.S. Stimulus analysing mechanisms. In Mechanisation of thought processes, (Vol 2), London: Her Majesty's Stationery Office, 1959. Cited inS. Kornblum (Ed.), Attention and performance IV, New York: Academic Press, 1973. Swets, J.A. Signal detection theory applied to vigilance. In R.R. Mackie (Ed.), Attention and performance IV, New York: Academic Press, 1973. Thomas, E.A. On expectancy and the speed and accuracy of responses. InS. Kornblum (Ed.), Attention and permance ]Y, New York: Academic Press, 1973. --Treisman, A.M., & Geffen, G. Selective attention and cerebral dominance in perceiving and responding to speech messages. Quarterly Journal of Experimental Psychology, 1968, 20, 139-151. Treisman, A.M.,-& Riley, J.G. Is selective attention selective perception or selective response? A further test. Journal of Exnerimental Psycholo~r, 1969, 12_, 27-34. Treisman, A.M., & Davies, A. Divided attention to ear and eye. InS. Kornblum (Ed.), Attention and performance IV, New York: Academic Press, 1973. Tulving, E., & Lindsay, P.H. Identification of simultaneously presented simple visual and auditory stimuli. In A.F. Sanders (Ed.), Attention and performance I, 37 Amsterdam: Humanities Press, 1970. Tyler, D.M., 1iJaag, W.L., &Holcomb, C.G. rllonitoringperformance across sense modes. An individual differences approach. Human Factors, 1972, 14(6), 539-547. Webber, C.E., & Adams, J.A. Effects of visual display mode on six hours of visual monitoring. Human Factors, 1964, 6, 13-20. Wilkinson, R. T., Morlock, H. C., & lrJilliams, H. L. Evoked cortical responses during vigilance. Psychonomic Science, 1966, 4, 221-222. Yntema, D.B., & Schulman, G.M. Response selection in keeping track of several things at once. InS. Kornblum (Ed.), Attention and Qerformance I, Amsterdam: ·Humanities Press, 1970.