International Journal of Psychophysiology, 13 (1992) 111-117
I 11
© 1992 Elsevier Science Publishers B.V. All rights reserved 0167-8760/92/$05.00 INTPSY 00403
Effect of task difficulty and interstimulus interval on blink parameters R o b e r t Goldstein a, Lance O. B a u e r b and John A. Stern a a Washington Unicersity Behacior Research Laboratory, Washington University, St. Louis, MO (USA) and b Unicersity of Connecticut Health Center, Department of Psychiatry, Farmington, CT (USA) (Accepted 18 June 1992)
Key words: Cognition; Task difficulty; Blink; Blink rate; Blink duration; Blink latency; RT; ISI
The effects of task difficulty and interstimulus interval (ISI) on blink rate, blink latency and blink duration, were studied in a modified Sternberg memory task in which either two or six characters were to be memorized. Stimuli were presented in ISI blocks at either 5.3 or 9.3 s (SOAs of 6 or 10 s). While b~,i,'lkrate and blink duration declined prior to each stimulus, the difficulty of the expected task (the length of the memory set) did not affect the rate of decline or the final prestimulus level. Concerning ISI, blink rate declined more rapidly during shorter ISis but the final prestimulus level, as was the case with task difficulty, was unaffected by the ISI duration. Presentation of the 6-character memory set produced a marked immediate inhibition of blinking. The data suggest that the encoding of visual stimuli is more akin to processes invoked in preparation for input than to ensuing processing stages since both encoding as well as preparation are accompanied by inhibition of blinking.
INTRODUCTION In a previous experiment (Bauer et al., 1987), eye blinks, among other variables, were recorded while subjects performed a modified Sternberg (1966) memory task. In that task, each trial consisted of three stimuli: a Cue, a Memory (set) and a Test stimulus. These were presented at 5.3 s ISis, which was also the intertrial interval. The Cue stimulus indicated the number of letters (1, 3 or 5) to appear in the upcoming Memory set. The latter, in turn, was followed by the Test stimulus, a single letter. The subject indicated, by a joystick response, whether the Test stimulus was or was not a member of the Memory set. Set size did not affect blink frequency in the Cue period There was simply a gradual reduction
Correspondence to: R. Goldstein, Department of Psychology, Washington University, St. Louis, MO 63130, USA.
in blink rate through the period prior to the Memory set, an anticipatory effect reported also by Bernstein et al. (1985). However, there was a dramatic blink inhibition at the outset of the Memory period following presentation of only the 5-letter set. This inhibition was followed immediately by a substantial rebound that reversed the set size relationship; blink rate for the large set was significantly higher than that for the smaller sets. This set size difference was significant throughout the Memory period. It was hypothesized that the prolonged blink inhibition following presentation of the longer Memory set was an indication of the additional time required for encoding that set. This view was consistent with previous data (Goldstein et al., 1985). It was possible, nevertheless, that this prolonged relative elevation of blink rate for the larger set could be due to the anticipation of the Test stimulus and not simply to a simple residual of the prior elevation or of continuing processing of the memory
112
set. This was tested here by including a condition with a long interstimulus interval. Further, to confirm the set size effect in the Memory period, maximum set size was increased to six, a value close to the upper limit a subject could perceive at the fixed exposure time required by the present procedure.
METHOD
for 700 ms at the center of the display. The Memory and Test stimuli were also centrally presented for 700 ms. Memory set letters were selected at random from 18 upper-case consonants (excluding 'Y', 'W', and 'V') and arranged in a series with three restrictions: in each set of 168 trials, the two set sizes occurred equally often, the Test stimulus was one of the Memory set letters on half the trials, and on such "match' trials it occurred with equal frequency at each position in the set.
Subjects Fourteen right-handed male Washington University students, ages 18-26, were paid for participating in this experiment. All had normal or corrected-to-normal vision.
Apparatus Sessions were run with subjects seated in a 2.3 m x 2.75 m electrically-shielded, sound-attenuated, room. Illumination was provided by two overhead incandescent lights on either side and slightly behind the subject. Ambient light intensity was 0.8 candela. A PDP 11/23 computer controlled stimulus presentation by activating a dot-matrix display unit (IEE Inc., 1 × 20 Dot Matrix Display Module #3600-14-020) centered behind a 1.3 cm wide transparent horizontal window in a sheet of black opaque plastic. The subject sat 1.5 m from the display with his right hand on a joystick. Electrooculographic (EOG) signals were recorded throughout the experiment. E O G was taken from a pair of miniature biopotential electrodes applied immediately above the eyebrow and below the left eye. The subject was grounded by an electrode applied to the center of his forehead. Interelectrode impedance was kept below 5 kohms. EOG was amplified by an automatic rezeroing amplifier (gain: 1.5 K, iowpass: DC to 1 kHz). Signals were stored digitally on computer disc (sampling rate: 200 Hz) for off-line analysis.
Stimuli The Cue stimuli were the numerals '2' or '6' (mean luminance = 8.21 cd/m2; approximate retinal angle was 15' wide x 20' high) projected
Procedure A trial consisted of three events occurring in sequence: a Cue stimulus, the Memory set and the Test stimulus. Interstimulus intervals for all stimuli in a given trial as well as the intertrial interval for that block was either 5.3 or 9.3 s Thus, S P A s (IS1 plus stimulus duration, 700 ms) were 6 and 10 s. The experiment was divided into 4 blocks of 84 trials, two on each of 2 days. In each 84 trial block, all conditions of the experiment, viz., set size, membership (i.e., whether or not the Test stimulus was a member of the Memory set), were represented equally. Each of the 4 blocks of 84 trials was subdivided into 4 sub-blocks of 21 trials. In two of these, ISI was 5.3 s, and in two, ISI was 9.3 s. The ISI was announced on the display prior to each sub-block. Instructions to subjects were similar to those in Bauer et al. (1987), but included an explanation of the ISI value. They were told the meaning of the Cue stimulus, and that the character set was to be memorized. They were instructed to make a rapid joystick response when the Test letter was presented indicating whether or not that letter was in the Memory set. For half the subjects~ 'yes' was a right joystick movement and 'no' was a left movement. For the other half, this was reversed.
DATA TREATMENT
Blink rate Blink rate was calculated for each 1-s period ('bins') throughout each trial. The number of
113
were adjusted using the conservative GeisserGreenhouse (1958) adjustment.
blinks in a given 1-s time bin was summed across trials of a block, divided by the total number of trials in the block (producing blinks per s), and multiplied by 60, conve~ing it to blinks/min.
Blinks Blink rate. The blink data, displayed in Fig. 1, were subjected to two ANOVAs: the first included the first six bins for both ISis, and the second, the remaining four bins (7-10) of the long ISI trials. All effects of interest appeared to be covered by these analyses. In the trust ANOVA, two main effects were significant: Bin (F(2.8,37.0) -- 18.56, P < 0.001), and Period (F(1.6,21.5)= 8.78, P = 0.003). The source of the Bin effect is clear: in general, blink rate declines from a high (25.2/rain) at the outset of the period to a low ( 1 5 . 3 / m i n ) in Bin 6. Equally clear is that the rate change over bins depends jointly on Period and Setsize (Bin × Period × Setsizc interaction F(2.5,32.6) = 12.33, P--0.001). Rate declined monotonically in all cases except in the Memory period, where the large set size started from a low in bin 1, increased to a high in bin 2, and then declined as in the other periods. In post-hoe analyses, this pattern produced a significant Bin × Setsize interac-
Blink latency The latency of a blink was defined as the period from the onset of the stimulus, of which there were three per trial, to the initiation of the first blink. Blink duration The duration of a blink was defined as extending from the pcint when the F.OG signal attained half of its maximum amplitude during eye closing, to the point when the signal crosses the same voltage level in the reopening phase.
RESULTS To simplify exposition, only those effects achieving significance ( a = 0.05) will be reported except in cases where a nonsignificant result would be of some relevance. Degrees of freedom
TESTPERIOD
N~EMORY PERIOD
CUE PERIOD
iUNKgATE(BLINK$/~INUTE)
|LINI[ R&T[(|LINK$/MINUT|)
IlLm[ 2ATE CO~.L~I:SI~,I~;U~'E)
~5I
O SHOIET~I. ~*;TIE/A rl SHOIITaL 6,.ITIE~ • IONG ~L 2-I,[~
I
i
B
11 LONGISL6-1TE,~ 30
~O
" ~ "
C
' x
,,'., =o.!
Io
2o.
\\
%
§IN N O .
BIN NO.
Fig. 1. Blinkrate as d functionof period, set size, ISI and Umebin.
BIN N O
..,° ..\.
114 TABLE 1
tion in the Memory period (F(2.13,27,71) = 14.91, P < 0.0001). For all bins, with the exception of the sixth (discussed below), differences due to set size were significant (all P values < 0.03). Finally, the Bin main effect was qualified by an interaction with ISI (F(2.4,31.1)=7.83, P = 0.001) with the larger set showing a slower rate of decline. In post-hoc analyses, ISI differences were significant for the fifth bin in the Memory period and for the sixth bin in both Cue and Memory periods (all P values < 0.003. Analysis of the final four bins of the long ISi produced a single significant effect, the main effect for Bin (F(2.2,28.2) = 10.12, P = 0.001); the rate decline initiated in the first six bins simply continued, undifferentiated by set size. Consistent with the ISl differences described above, an analysis of rate in the last bin only (Bin 6 for the short ISis and Bin 10 for the long ISis) yielded no significant main or interaction effects. Blink rate, though decreasing at different rates for different ISis, achieved the same low level at the end ot their respective periods.
CUE PERIOD
Blink latency as a fimction o f period and set size Set size
Cue
Mere
Test
Two Six
772 800
865 1039
874 865
Blink latency. Blink latency, presented in Table I, was subjected to a 2(ISl)x 2(Match/Mismatch) x 2(Setsize) x 3(Period) ANOVA. Two main effects were significant: Setsize (F(1,13)= 7.29, P = 0.018) and Period (F(2.0, 25.5)= 6.52, P = 0.005). More important, there was a significant Setsize x Period interaction (F(1.26,16.33) = 4.36, P = 0.045). The Setsize effect was restricted to the Memory period: the larger set produced a significantly longer blink latency than the short set (Setsize at Memory F(1,13)= 6.11, P = 0.028). The Period effect, a longer latency in the Memory period, was simply another reflection of the long latency for the 6-item set in that period.
MEMORY PERIOD
|LINg 0LIIAItOM (J~||C)
TEST PERIOD
tLh~l[ 0U|ATION [ M | | ¢ )
§LI~[ DU~ATSON¢M~£C)
la0 O SNOIT151.]-IrE,~.......,..~ ! •
tONG ~L ~,tIIA
U iO~40I$L6-1TIM I/0" ~
llO
1601 ~ 0 I + I
i
",~. 0
\
\ \ \
IS0.
150 -
....
-
140
I 110
.
2
.
.
3
.
4
.
.
S
ml i .
6
8ZHNO.
7
8
.
I
9
10
\
"e--o\."
°.
1,10
rio
i|0
i
,
,
3
4
+IH H O .
Fig. 2. Blink duration as a function of period, set size, ISi and time bin.
,
,
S 6 BtN NO.
7
8
9
I0
115
Blink duration. Bin-by-bin blink duration data are displayed in Fig. 2. The blink duration data were treated in the same manner as the rate data: with two ANOVAs. Significant main effects in the first ANOVA were: ISI (F(1,12)= 26.69, P = 0 . 0 0 0 2 ) , Bin (F(|.6,19.2)= 13.0, P ffi0.0005), and Period (F(1.9,23.0) = 19.51, P < 0.0001). The ISI effect indicates that the duration of blinks in the first 6 s was shorter when ISI was 5.3 s than when it was 9.3 s. A similar trend, albeit not signifieant, was observed for blink rate. There was no interaction of ISI and Setsize for blink duration. The significant Bin effect supports the observation that blink duration declines within periods, as does rate. As for the Period effect, blink duration increased across periods within a trial. In the second ANOVA, the Period effect of the first ANOVA was sustained (F(1.5,18.4)-8.56, P = 0.004). The Setsize x Period interaction was also significant (F(2.0,24.0)= 4,79, P < 0.018). Though the source of this interaction appears to lie in the difference between set sizes in the Memory period (duration for the 6-letter set was longer than for the 2-letter set) as compared to the differences in the Cue and Test periods (note the inversion in the Cue Period), none of the paired comparisons of Setsize for the three periods was significant. Finally, blink duration in the final bin of the t~ro ISI conditions, as was also the case for blink rate, did not differ (mean for the short ISI was 138 ms and for the long ISI, was 139 ms). Performance Reaction time. A median RT was obtained for each subject in each experimental condition. Tri-
TABLE III Error rate (percent incorrect) as a function of ISI and set size Set size
Two Six
ISI (s) six
ten
5,93 9.92
4.03 13.24
als on which incorrect responses were made were excluded from the RT analysis. A 2(ISI) × 2(Setsize)) × 2(Match/mismatch) repeated measures ANOVA was performed on the median RTs (Table II), which demonstrated, for both positive and negative judgments, that there was a nonsignificant (F(1,13)= 4.51, P = 0.054) increase (111 ms) in RT as a function of set size. The Match/mismatch factor was significant (F(1,13) = 29.6, P --- 0.0001). Reaction times were found to be 202.5 ms longer for correct negative (mismatch)judgments than for correct positive judgments. Response accuracy. The effects of ISI, Setsize, and Match/Mismatch on error rate (arcsinetransformed) were examined in a three-way ANOVA. The data are presented in Table III. Though Setsize did not have an effect on reaction time (see above), Setsize was manifested in a significantly higher error rate for the large set (F(1,13) = 15.7, P < 0.002). Further, there was a significant interaction of Setsize with ISI (F(1,13) = 9.21, P < 0.01); the larger set produced disproportionately more errors at the longer ISI. Whereas mismatch judgments took longer than match judgments, as reported above, the tradeoff was that fewer errors were made in mismatch judgments (F(1,13) = 5.4, P = 0.037).
DISCUSSION TABLE 1I Reaction time (ms) as a function of membership (Match/ mismatch) and set size Set size
Match
Mismatch
Two Six
893 1009
1 101 1206
Increasing the ISI had the desired clarifying effect on the relationship between the underlying cognitive processes and our eyeblink measures. As in the Bauer et al. (1987) study, set size did not affect the blink pattern in the Cue period. The relationship between blink rate and experi-
116
mental condition is consistent with the view that the blink rate reflects cognitive demands. Thus, blink rates were high at the outset of the Cue and Test periods as they were for the 2-item set in the Memory period. In these instances, only one or two characters had to be processed. In contrast, there was a significant delay in blinking following presentation of the large memory set. This theme fits in well with the results of the Bauer et al. (1987) study, where neither the 1- or 3-item sets produced blink inhibition. In that study, only the five item set produced the inhibitory effect. Summing up the Memory period results for the two studies, one can discern a transition from no inhibition for the 1-, 2- or 3-item sets, (all means approx. 30 bpm) through partial inhibition for the 5-item set (mean = approx. 18 bpm), and finally to substantial inhibition for the 6-item set (mean = approx. 14 bpm). Apparently, memorizing five or six characters is a significant challenge to a subject's capacity, one that is ~e,~.ted in blink inhibition. Following the initial depression in blink rate for the 6-item set in the Memory period, the subsequent increase may be interpreted as indicating the conclusion of the encoding process. The effect in the present study was even more accentuated than in the previous study• Blink rate for the 6-item set, while converging with that for the smaller set later in the Memory period, was otherwise higher throughout the period, as in the earlier study. This persistent set size difference in blink rate, may suggest that stimulus processing (rehearsal?) for the large set is continuing beyond the encoding stage, at least to a greater extent than in the small set. In the short ISI condition, the differences disappeared by the end of the Memory period. And in the long !SI condition, there were no set size differences in any bin. Relevant to our initial concerns, this would indicate that the high blink rate in the larger set (after the initial inhibition) are related to preceding processes in this period and not to the anticipation of demands of the Test period. The pattern of blink latency observed in the earlier study was repeated here. The latency measure gives us a more detailed picture of the blink inhibition following the large set. The greater
attentional demands made by presentation of the 6-, as compared to the 2-, item Memory set is reflected in an increase in the later, cy to the first blink and, consequently, a lower blank rate in the first bin f o r the larger set. An interesting addendum to the blink latency observations is the absence, in both studies, of any suggestion of a set size effect in response to the Test stimulus. This contrasts sharply with the inhibition noted at the beginning of the Memory period. We conclude that blink inhibition occurs only when the underlying processes consume a substantial period of time. The encoding of a significant chunk of information meets this criterion. On the other hand, from an empirical standpoint, the criterion is not met by either the enc,,n~,,~, of the Test stimulus (apparently requirin~ less time than the 2-item memory set) or the process by which the Test stimulus is compared with the memorized set. Prolonging the poststimulus processing by altering tee nature of the task would test this proposition. In contrast to the reduction in the rate of blinking for the large set at the outset of the Memory period, blink duration did not show the initial Memory period depression. Those blinks that occurred were the longest in duration of any other period• Nevertheless, the reduction in blink duration in anticipation of task stimufi supports the conclusion that this measure reflects the degree of attention to external stimulation and, consequently, carries information that is not redundant with the rate measure. That is, though both blink rate and blink duration reflect the anticipation of an imperative stimulus, only rate is affected by the actual encoding of that stimulus.
ACKNOWLEDGMENTS This research was conducted under Air Force Office of Scientific Research, contract F4962083-C-0059, and the Air Force Office of Aerospace Medicine, contract F33615-87-C-06-03. We would like to thank the Project Officers, Alfred Fregly, Ph.D., and William Storm, Ph.D., for their support and encouragement.
117 REFERENCES Bauer, L.O., Go]dste/n, R. and Stem, LA. (1q87) Effects of information prcg-essing demands on physiotogical response palteru~ Human Factors, 29: 213-234. Be~steh~, A.S., Taylor, K.W., Weh~stch% E. aad R/edc|, J. (1985) The effeclt of st~t~uius significance on re|ative|y s~sla~ined (mnic-|ike) and re~ative|y warEs~ent (pkasie-~ike) ~sg~e~ of e|ec~r~erm:d, heart rate ~ d eyeblL~k respoz~s. /~4. P~xr&o#.. 21: 183-228.
GeLssen S. and Greenhouse, S.W. (1958) An extension of Box's results on the use of the F distrt~0ution in multivariate statistics. Amtal. Mathemat. Stat., 29: 885-891. Gold.stein, R., Walrath, LC., Stern, J.S. and Strock, B.D. (t985) B~imk activi+ +n a discr/mination task as a function of stimulus mod~ity and schedule of presentation. Psydrophysiotagy, 22: 629--635. Steraberg, S+ (t966) High speed ~ r m i n g ha human memory. Science, 153: 652-654.
I 11
© 1992 Elsevier Science Publishers B.V. All rights reserved 0167-8760/92/$05.00 INTPSY 00403
Effect of task difficulty and interstimulus interval on blink parameters R o b e r t Goldstein a, Lance O. B a u e r b and John A. Stern a a Washington Unicersity Behacior Research Laboratory, Washington University, St. Louis, MO (USA) and b Unicersity of Connecticut Health Center, Department of Psychiatry, Farmington, CT (USA) (Accepted 18 June 1992)
Key words: Cognition; Task difficulty; Blink; Blink rate; Blink duration; Blink latency; RT; ISI
The effects of task difficulty and interstimulus interval (ISI) on blink rate, blink latency and blink duration, were studied in a modified Sternberg memory task in which either two or six characters were to be memorized. Stimuli were presented in ISI blocks at either 5.3 or 9.3 s (SOAs of 6 or 10 s). While b~,i,'lkrate and blink duration declined prior to each stimulus, the difficulty of the expected task (the length of the memory set) did not affect the rate of decline or the final prestimulus level. Concerning ISI, blink rate declined more rapidly during shorter ISis but the final prestimulus level, as was the case with task difficulty, was unaffected by the ISI duration. Presentation of the 6-character memory set produced a marked immediate inhibition of blinking. The data suggest that the encoding of visual stimuli is more akin to processes invoked in preparation for input than to ensuing processing stages since both encoding as well as preparation are accompanied by inhibition of blinking.
INTRODUCTION In a previous experiment (Bauer et al., 1987), eye blinks, among other variables, were recorded while subjects performed a modified Sternberg (1966) memory task. In that task, each trial consisted of three stimuli: a Cue, a Memory (set) and a Test stimulus. These were presented at 5.3 s ISis, which was also the intertrial interval. The Cue stimulus indicated the number of letters (1, 3 or 5) to appear in the upcoming Memory set. The latter, in turn, was followed by the Test stimulus, a single letter. The subject indicated, by a joystick response, whether the Test stimulus was or was not a member of the Memory set. Set size did not affect blink frequency in the Cue period There was simply a gradual reduction
Correspondence to: R. Goldstein, Department of Psychology, Washington University, St. Louis, MO 63130, USA.
in blink rate through the period prior to the Memory set, an anticipatory effect reported also by Bernstein et al. (1985). However, there was a dramatic blink inhibition at the outset of the Memory period following presentation of only the 5-letter set. This inhibition was followed immediately by a substantial rebound that reversed the set size relationship; blink rate for the large set was significantly higher than that for the smaller sets. This set size difference was significant throughout the Memory period. It was hypothesized that the prolonged blink inhibition following presentation of the longer Memory set was an indication of the additional time required for encoding that set. This view was consistent with previous data (Goldstein et al., 1985). It was possible, nevertheless, that this prolonged relative elevation of blink rate for the larger set could be due to the anticipation of the Test stimulus and not simply to a simple residual of the prior elevation or of continuing processing of the memory
112
set. This was tested here by including a condition with a long interstimulus interval. Further, to confirm the set size effect in the Memory period, maximum set size was increased to six, a value close to the upper limit a subject could perceive at the fixed exposure time required by the present procedure.
METHOD
for 700 ms at the center of the display. The Memory and Test stimuli were also centrally presented for 700 ms. Memory set letters were selected at random from 18 upper-case consonants (excluding 'Y', 'W', and 'V') and arranged in a series with three restrictions: in each set of 168 trials, the two set sizes occurred equally often, the Test stimulus was one of the Memory set letters on half the trials, and on such "match' trials it occurred with equal frequency at each position in the set.
Subjects Fourteen right-handed male Washington University students, ages 18-26, were paid for participating in this experiment. All had normal or corrected-to-normal vision.
Apparatus Sessions were run with subjects seated in a 2.3 m x 2.75 m electrically-shielded, sound-attenuated, room. Illumination was provided by two overhead incandescent lights on either side and slightly behind the subject. Ambient light intensity was 0.8 candela. A PDP 11/23 computer controlled stimulus presentation by activating a dot-matrix display unit (IEE Inc., 1 × 20 Dot Matrix Display Module #3600-14-020) centered behind a 1.3 cm wide transparent horizontal window in a sheet of black opaque plastic. The subject sat 1.5 m from the display with his right hand on a joystick. Electrooculographic (EOG) signals were recorded throughout the experiment. E O G was taken from a pair of miniature biopotential electrodes applied immediately above the eyebrow and below the left eye. The subject was grounded by an electrode applied to the center of his forehead. Interelectrode impedance was kept below 5 kohms. EOG was amplified by an automatic rezeroing amplifier (gain: 1.5 K, iowpass: DC to 1 kHz). Signals were stored digitally on computer disc (sampling rate: 200 Hz) for off-line analysis.
Stimuli The Cue stimuli were the numerals '2' or '6' (mean luminance = 8.21 cd/m2; approximate retinal angle was 15' wide x 20' high) projected
Procedure A trial consisted of three events occurring in sequence: a Cue stimulus, the Memory set and the Test stimulus. Interstimulus intervals for all stimuli in a given trial as well as the intertrial interval for that block was either 5.3 or 9.3 s Thus, S P A s (IS1 plus stimulus duration, 700 ms) were 6 and 10 s. The experiment was divided into 4 blocks of 84 trials, two on each of 2 days. In each 84 trial block, all conditions of the experiment, viz., set size, membership (i.e., whether or not the Test stimulus was a member of the Memory set), were represented equally. Each of the 4 blocks of 84 trials was subdivided into 4 sub-blocks of 21 trials. In two of these, ISI was 5.3 s, and in two, ISI was 9.3 s. The ISI was announced on the display prior to each sub-block. Instructions to subjects were similar to those in Bauer et al. (1987), but included an explanation of the ISI value. They were told the meaning of the Cue stimulus, and that the character set was to be memorized. They were instructed to make a rapid joystick response when the Test letter was presented indicating whether or not that letter was in the Memory set. For half the subjects~ 'yes' was a right joystick movement and 'no' was a left movement. For the other half, this was reversed.
DATA TREATMENT
Blink rate Blink rate was calculated for each 1-s period ('bins') throughout each trial. The number of
113
were adjusted using the conservative GeisserGreenhouse (1958) adjustment.
blinks in a given 1-s time bin was summed across trials of a block, divided by the total number of trials in the block (producing blinks per s), and multiplied by 60, conve~ing it to blinks/min.
Blinks Blink rate. The blink data, displayed in Fig. 1, were subjected to two ANOVAs: the first included the first six bins for both ISis, and the second, the remaining four bins (7-10) of the long ISI trials. All effects of interest appeared to be covered by these analyses. In the trust ANOVA, two main effects were significant: Bin (F(2.8,37.0) -- 18.56, P < 0.001), and Period (F(1.6,21.5)= 8.78, P = 0.003). The source of the Bin effect is clear: in general, blink rate declines from a high (25.2/rain) at the outset of the period to a low ( 1 5 . 3 / m i n ) in Bin 6. Equally clear is that the rate change over bins depends jointly on Period and Setsize (Bin × Period × Setsizc interaction F(2.5,32.6) = 12.33, P--0.001). Rate declined monotonically in all cases except in the Memory period, where the large set size started from a low in bin 1, increased to a high in bin 2, and then declined as in the other periods. In post-hoe analyses, this pattern produced a significant Bin × Setsize interac-
Blink latency The latency of a blink was defined as the period from the onset of the stimulus, of which there were three per trial, to the initiation of the first blink. Blink duration The duration of a blink was defined as extending from the pcint when the F.OG signal attained half of its maximum amplitude during eye closing, to the point when the signal crosses the same voltage level in the reopening phase.
RESULTS To simplify exposition, only those effects achieving significance ( a = 0.05) will be reported except in cases where a nonsignificant result would be of some relevance. Degrees of freedom
TESTPERIOD
N~EMORY PERIOD
CUE PERIOD
iUNKgATE(BLINK$/~INUTE)
|LINI[ R&T[(|LINK$/MINUT|)
IlLm[ 2ATE CO~.L~I:SI~,I~;U~'E)
~5I
O SHOIET~I. ~*;TIE/A rl SHOIITaL 6,.ITIE~ • IONG ~L 2-I,[~
I
i
B
11 LONGISL6-1TE,~ 30
~O
" ~ "
C
' x
,,'., =o.!
Io
2o.
\\
%
§IN N O .
BIN NO.
Fig. 1. Blinkrate as d functionof period, set size, ISI and Umebin.
BIN N O
..,° ..\.
114 TABLE 1
tion in the Memory period (F(2.13,27,71) = 14.91, P < 0.0001). For all bins, with the exception of the sixth (discussed below), differences due to set size were significant (all P values < 0.03). Finally, the Bin main effect was qualified by an interaction with ISI (F(2.4,31.1)=7.83, P = 0.001) with the larger set showing a slower rate of decline. In post-hoc analyses, ISI differences were significant for the fifth bin in the Memory period and for the sixth bin in both Cue and Memory periods (all P values < 0.003. Analysis of the final four bins of the long ISi produced a single significant effect, the main effect for Bin (F(2.2,28.2) = 10.12, P = 0.001); the rate decline initiated in the first six bins simply continued, undifferentiated by set size. Consistent with the ISl differences described above, an analysis of rate in the last bin only (Bin 6 for the short ISis and Bin 10 for the long ISis) yielded no significant main or interaction effects. Blink rate, though decreasing at different rates for different ISis, achieved the same low level at the end ot their respective periods.
CUE PERIOD
Blink latency as a fimction o f period and set size Set size
Cue
Mere
Test
Two Six
772 800
865 1039
874 865
Blink latency. Blink latency, presented in Table I, was subjected to a 2(ISl)x 2(Match/Mismatch) x 2(Setsize) x 3(Period) ANOVA. Two main effects were significant: Setsize (F(1,13)= 7.29, P = 0.018) and Period (F(2.0, 25.5)= 6.52, P = 0.005). More important, there was a significant Setsize x Period interaction (F(1.26,16.33) = 4.36, P = 0.045). The Setsize effect was restricted to the Memory period: the larger set produced a significantly longer blink latency than the short set (Setsize at Memory F(1,13)= 6.11, P = 0.028). The Period effect, a longer latency in the Memory period, was simply another reflection of the long latency for the 6-item set in that period.
MEMORY PERIOD
|LINg 0LIIAItOM (J~||C)
TEST PERIOD
tLh~l[ 0U|ATION [ M | | ¢ )
§LI~[ DU~ATSON¢M~£C)
la0 O SNOIT151.]-IrE,~.......,..~ ! •
tONG ~L ~,tIIA
U iO~40I$L6-1TIM I/0" ~
llO
1601 ~ 0 I + I
i
",~. 0
\
\ \ \
IS0.
150 -
....
-
140
I 110
.
2
.
.
3
.
4
.
.
S
ml i .
6
8ZHNO.
7
8
.
I
9
10
\
"e--o\."
°.
1,10
rio
i|0
i
,
,
3
4
+IH H O .
Fig. 2. Blink duration as a function of period, set size, ISi and time bin.
,
,
S 6 BtN NO.
7
8
9
I0
115
Blink duration. Bin-by-bin blink duration data are displayed in Fig. 2. The blink duration data were treated in the same manner as the rate data: with two ANOVAs. Significant main effects in the first ANOVA were: ISI (F(1,12)= 26.69, P = 0 . 0 0 0 2 ) , Bin (F(|.6,19.2)= 13.0, P ffi0.0005), and Period (F(1.9,23.0) = 19.51, P < 0.0001). The ISI effect indicates that the duration of blinks in the first 6 s was shorter when ISI was 5.3 s than when it was 9.3 s. A similar trend, albeit not signifieant, was observed for blink rate. There was no interaction of ISI and Setsize for blink duration. The significant Bin effect supports the observation that blink duration declines within periods, as does rate. As for the Period effect, blink duration increased across periods within a trial. In the second ANOVA, the Period effect of the first ANOVA was sustained (F(1.5,18.4)-8.56, P = 0.004). The Setsize x Period interaction was also significant (F(2.0,24.0)= 4,79, P < 0.018). Though the source of this interaction appears to lie in the difference between set sizes in the Memory period (duration for the 6-letter set was longer than for the 2-letter set) as compared to the differences in the Cue and Test periods (note the inversion in the Cue Period), none of the paired comparisons of Setsize for the three periods was significant. Finally, blink duration in the final bin of the t~ro ISI conditions, as was also the case for blink rate, did not differ (mean for the short ISI was 138 ms and for the long ISI, was 139 ms). Performance Reaction time. A median RT was obtained for each subject in each experimental condition. Tri-
TABLE III Error rate (percent incorrect) as a function of ISI and set size Set size
Two Six
ISI (s) six
ten
5,93 9.92
4.03 13.24
als on which incorrect responses were made were excluded from the RT analysis. A 2(ISI) × 2(Setsize)) × 2(Match/mismatch) repeated measures ANOVA was performed on the median RTs (Table II), which demonstrated, for both positive and negative judgments, that there was a nonsignificant (F(1,13)= 4.51, P = 0.054) increase (111 ms) in RT as a function of set size. The Match/mismatch factor was significant (F(1,13) = 29.6, P --- 0.0001). Reaction times were found to be 202.5 ms longer for correct negative (mismatch)judgments than for correct positive judgments. Response accuracy. The effects of ISI, Setsize, and Match/Mismatch on error rate (arcsinetransformed) were examined in a three-way ANOVA. The data are presented in Table III. Though Setsize did not have an effect on reaction time (see above), Setsize was manifested in a significantly higher error rate for the large set (F(1,13) = 15.7, P < 0.002). Further, there was a significant interaction of Setsize with ISI (F(1,13) = 9.21, P < 0.01); the larger set produced disproportionately more errors at the longer ISI. Whereas mismatch judgments took longer than match judgments, as reported above, the tradeoff was that fewer errors were made in mismatch judgments (F(1,13) = 5.4, P = 0.037).
DISCUSSION TABLE 1I Reaction time (ms) as a function of membership (Match/ mismatch) and set size Set size
Match
Mismatch
Two Six
893 1009
1 101 1206
Increasing the ISI had the desired clarifying effect on the relationship between the underlying cognitive processes and our eyeblink measures. As in the Bauer et al. (1987) study, set size did not affect the blink pattern in the Cue period. The relationship between blink rate and experi-
116
mental condition is consistent with the view that the blink rate reflects cognitive demands. Thus, blink rates were high at the outset of the Cue and Test periods as they were for the 2-item set in the Memory period. In these instances, only one or two characters had to be processed. In contrast, there was a significant delay in blinking following presentation of the large memory set. This theme fits in well with the results of the Bauer et al. (1987) study, where neither the 1- or 3-item sets produced blink inhibition. In that study, only the five item set produced the inhibitory effect. Summing up the Memory period results for the two studies, one can discern a transition from no inhibition for the 1-, 2- or 3-item sets, (all means approx. 30 bpm) through partial inhibition for the 5-item set (mean = approx. 18 bpm), and finally to substantial inhibition for the 6-item set (mean = approx. 14 bpm). Apparently, memorizing five or six characters is a significant challenge to a subject's capacity, one that is ~e,~.ted in blink inhibition. Following the initial depression in blink rate for the 6-item set in the Memory period, the subsequent increase may be interpreted as indicating the conclusion of the encoding process. The effect in the present study was even more accentuated than in the previous study• Blink rate for the 6-item set, while converging with that for the smaller set later in the Memory period, was otherwise higher throughout the period, as in the earlier study. This persistent set size difference in blink rate, may suggest that stimulus processing (rehearsal?) for the large set is continuing beyond the encoding stage, at least to a greater extent than in the small set. In the short ISI condition, the differences disappeared by the end of the Memory period. And in the long !SI condition, there were no set size differences in any bin. Relevant to our initial concerns, this would indicate that the high blink rate in the larger set (after the initial inhibition) are related to preceding processes in this period and not to the anticipation of demands of the Test period. The pattern of blink latency observed in the earlier study was repeated here. The latency measure gives us a more detailed picture of the blink inhibition following the large set. The greater
attentional demands made by presentation of the 6-, as compared to the 2-, item Memory set is reflected in an increase in the later, cy to the first blink and, consequently, a lower blank rate in the first bin f o r the larger set. An interesting addendum to the blink latency observations is the absence, in both studies, of any suggestion of a set size effect in response to the Test stimulus. This contrasts sharply with the inhibition noted at the beginning of the Memory period. We conclude that blink inhibition occurs only when the underlying processes consume a substantial period of time. The encoding of a significant chunk of information meets this criterion. On the other hand, from an empirical standpoint, the criterion is not met by either the enc,,n~,,~, of the Test stimulus (apparently requirin~ less time than the 2-item memory set) or the process by which the Test stimulus is compared with the memorized set. Prolonging the poststimulus processing by altering tee nature of the task would test this proposition. In contrast to the reduction in the rate of blinking for the large set at the outset of the Memory period, blink duration did not show the initial Memory period depression. Those blinks that occurred were the longest in duration of any other period• Nevertheless, the reduction in blink duration in anticipation of task stimufi supports the conclusion that this measure reflects the degree of attention to external stimulation and, consequently, carries information that is not redundant with the rate measure. That is, though both blink rate and blink duration reflect the anticipation of an imperative stimulus, only rate is affected by the actual encoding of that stimulus.
ACKNOWLEDGMENTS This research was conducted under Air Force Office of Scientific Research, contract F4962083-C-0059, and the Air Force Office of Aerospace Medicine, contract F33615-87-C-06-03. We would like to thank the Project Officers, Alfred Fregly, Ph.D., and William Storm, Ph.D., for their support and encouragement.
117 REFERENCES Bauer, L.O., Go]dste/n, R. and Stem, LA. (1q87) Effects of information prcg-essing demands on physiotogical response palteru~ Human Factors, 29: 213-234. Be~steh~, A.S., Taylor, K.W., Weh~stch% E. aad R/edc|, J. (1985) The effeclt of st~t~uius significance on re|ative|y s~sla~ined (mnic-|ike) and re~ative|y warEs~ent (pkasie-~ike) ~sg~e~ of e|ec~r~erm:d, heart rate ~ d eyeblL~k respoz~s. /~4. P~xr&o#.. 21: 183-228.
GeLssen S. and Greenhouse, S.W. (1958) An extension of Box's results on the use of the F distrt~0ution in multivariate statistics. Amtal. Mathemat. Stat., 29: 885-891. Gold.stein, R., Walrath, LC., Stern, J.S. and Strock, B.D. (t985) B~imk activi+ +n a discr/mination task as a function of stimulus mod~ity and schedule of presentation. Psydrophysiotagy, 22: 629--635. Steraberg, S+ (t966) High speed ~ r m i n g ha human memory. Science, 153: 652-654.
