Echoic Memory In Primitive Auditory Selective Attention
Ben Weedon & Zofia Kaminska
Psychology Depart m e n t, City University, London, EC1V 0HB email: [email protected]
Abstract
This study explored pos sible confou n di ng effects of echoic mem ory on estim ate s of atten tio nal capacity by attem p t i ng to estim ate capacity while controlling the effect of echoic mem o ry. Detection of a target stimulu s and identification of its carrier strea m was investigate d as a functio n of variation in nu m ber of concur re n t non - overlap ping auditory strea m s (1 to 4) and of condition of echoic mem o ry involveme n t (available or eliminate d via articulatory supp r e s sio n). Error rates were found to increase non - linearly as a functio n of numbe r of strea m s, but with a differen t point of discontin uity – indicative of a proces sing limitatio n – under different conditions of memory involveme n t, a higher capacity (3 as oppos e d to 2 stream s) being achieved with echoic mem ory contribu tio n. Detection respo n s e latencies also increase d as a function of nu mb e r of strea m s, but the increases were linear. Echoic mem o ry also significantly reduce d the respo n s e latency. These findings, which implicate echoic mem o ry as a contribut o ry factor in estimat e s of audito ry atten tio n, may help to resolve discre pa ncies in previous researc h and have implications for modeling auditory atten tion al proces s e s.
1 INTRODUCTION
Previous research into the proces si ng capacity of selective atten tio n has prod uce d conflicting results. Early findings (e.g., Broadben t, 1958; Treism a n, 1964 and Sloboda and Edworthy, 1981) sugges te d a serial mo de of proces sing in a figure / g r o u n d Gestaltian mann er, in which a single chann el receives full proces sin g, other s remaining as backgro u n d . More recent researc h (Huron, 1989; Gregory, 1990 and Ludlam, 1993) has discovere d what appear s to be a processing
capacity of 3 auditory strea m s, sugges ting a greater degree of parallel proces sin g.. A possible cause of these discrepa nt finding s may be that the meas ur e s of atten tio n al capacity in the latter experi me n t s relied explicitly on memory proces s es. For exam ple, a typical task used to estim ate atten tio n al capacity is to ask listener s to identify which of varying num be r of simulta ne o u s auditory strea m s contains a target stim ulu s. However, listeners may be able to perfor m this task not becaus e they are able to proces s all input s simulta n e o u s ly, but because they may be able to retrieve traces of strea m s unat te n d e d to at the mo me n t of percep tio n, using their memory. Thus, higher cognitive proces s e s are confou n di ng the meas u re of atten tio nal limitatio ns.
The aspect of memo ry most likely to be involved in auditory attentio n is echoic memory. Cowan (1984), sum m a ri sing research into auditory mem o ry, note d that there is evidence for two separa t e syste m s. Both are containe d within the pho nological short - term store as pro po s e d by Baddeley (1974). One syste m deals with elemen t a ry proces si ng of auditory stimuli and has a minut e tem po r al span (the short - term audito ry proces s, or ‘senso ry register’ in Treism a n’s mod el), whereas the other (the long - term audito ry store) has a longer span, a smaller spatial capacity and is synony m o u s with echoic mem o ry. It is echoic mem o ry which could have been utilised by participa nt s in studies reporting 3- strea m capacity (Huron, 1989; Gregory, 1990 and Ludlam, 1993).
To test this hypot he si s, the presen t study explored the effects of blocking participa n t s’ use of echoic memo ry on a task similar to those used previou sly to deter mi ne attention al capacity (Huron, 1989; Ludlam, 1993). The techniq ue used to preclude the use of echoic mem ory was articulato ry supp re s sio n. This has been shown to be a consiste n t met ho d of occupying the limited space of the phon ological loop, thereby preven ti ng subject s from using echoic memory to store new material and resulting in the decay of material already held (Baddeley, 1975; Salamé and Baddeley, 1982; Keller, Cowan and Saults, 1995). It involves participan t s prod ucing contin u o u s verbal out p u t of minimal cognitive load.
Thus this experime n t investigate d auditory attention capacity while atte m p ti ng to control the role of echoic memory.
2 METHOD
2.1 Design
Eighteen participa n t s took part in the experime n t. They were rand o m ly assigned to under ta k e the Echoic or Non - echoic condition first. The experime nt took the form of a four by two, com pletely within - participan t s, factorial design. The first variable, Strea m s, was the num b er of concurre n t auditory strea m s. This had four levels, with the num b e r of simulta n e o u s strea m s being 1, 2, 3 or 4.
The secon d variable, Memory, was memory involvem en t, with two levels, Echoic and Non - echoic, manip ulat e d within –participa nt s.
Under the Echoic conditio n participa n t s perfor m e d the target detection task witho u t any experime n t al blocking of memo ry involveme n t. In the Non - echoic condition echoic mem o ry contrib u tion was blocked by requiring the participa n t s to recite aloud rand o m ly generate d num be r s fro m 1 to 8 as they appeare d on the com pu t e r screen during the basic task. This served to sup pre s s any intern al auditory imagery activity.
Respon s e times to the target stim ulu s in each level of the experi me n t were record e d by the compu t e r, as were the percen tage of incorrect respo n s e s. The time limit within which participan t s were to respo n d was 5 secon d s. Incorrect respo n s e s were also given a score of 5 second s. So that participant s could get used to the different soun d s of the instru m e n t s and the dem a n d s of the experi me n t, and also that they might refine their reactio ns, a practice session was include d.
Participan t s underwe n t 160 trials in two blocks of 80, one block for each memo ry condition (Echoic / Non - echoic). Within each block of 80 there were 20 blocks for each level of Stream s, the prese nt atio n of which were rando m is e d.
2.2 Stimulus Materials
The stimuli were designe d to be 4 individu ally identifiable auditory strea m s when presen t e d concurre n tly, with no stimulu s being any more salient than any other when 2 or more were playing
simulta n e o u s ly. Each strea m consis te d of a ‘melody’ that lasted for a few secon d s before a target stimulu s was presen t e d as the final note in one of the strea m s. To achieve the struct u re of the strea m s
Bregma n’s (1990) guidelines on percept u al strea mi ng and Auditory Scene Analysis were followed.
The audito ry stimuli were four individual musical stream s varying in timbre by being generat e d by four different instru m e n t s (piano, whistle, harp sich or d and saxoph o n e) and also varying in their
“bright ne s s”, with the highes t pitched instru m e n t having the
brightes t soun d. Each instru m e n t played a chrom atic set of notes in different pitch ranges in a symm e t rical ‘v’ shape of a 3 - semito ne sprea d. The ‘melodies’ of the instru m e n t s are shown in Figure 1.
The different strea m s did not overlap with each other in pitch. When there was more than one instru m e n t playing note onset s were synch ro n o u s. The strea m s were presen t e d in mono at a rate of one note every 200 milliseco n d s. They were played repeate dly for between 5 and 7 second s (varying rand o m ly) until the target stimulu s was presen te d. The target stim ulu s replace d one of the notes in one of the strea m s, selecte d at rand o m. It consiste d of a burs t of white noise. The strea m s were represe n t e d on the com p u t e r VDU by their respective name s, written in blue, in 4 blue coloured boxes. When a strea m was playing its name and box turne d red.
Each box was ascribe d a nu m b er, fro m 1 to 4. Participan t s were instructe d to respo n d as quickly as pos sible to the target by pressin g the nu m ber key that corres p o n d e d to the strea m in which it was prese n t e d.
Figure 1 - Melodies of the Stimuli Instru m e n t s
2.3 Respon s e Measurem ents
Participan t s were rando m ly assigne d to undert a ke the Echoic or Non - echoic condition first. In the Non - echoic condition subject s were told to respo n d as quickly as pos sible to the target stimul u s by pressin g the key corres p o n d i n g to the musical instru m e n t in which the stim ulu s appeare d. In addition, they were instructe d to repeat aloud digits appearing in a rand o m order in a different portion of the screen. The import a n ce of maint aining accuracy on this latter task was emph a si se d. In the Echoic conditio n target detection instruction s were the same, but there was no rand o m digit sequen ce to occupy the subject s. The 80 test trials of each level of Memory followed a practise session for that level. When subject s were confide nt that they could perfor m the task, that level of the experi me n t began. Between each level of Memory there was a 2 minu te interval. Participa nt s were fully de - briefed at the end of testing.
RESULTS
Analysis of Errors
A 2- factor repeate d measu re s ANOVA was perfor m e d on the error data. There was a significant effect of num b e r of strea m s presen te d simulta n e o u s ly on the subjects’ ability to discern in which strea m the target stimulu s appeare d, with error rates increasing as a functio n of increasing nu mb e r s of strea m s (F 36, 2 = 196.9, p <
0.0005). Echoic mem o ry involveme n t also had a significant effect on ability to discern in which stream the target stimulu s occurre d (F 17, 1
= 40.83, p <0.00 05). Figure 2 shows that echoic memory usage led to an increase in correct resp o n s e s. There was also a significant interaction between Stream s and Memory (F 34, 2 = 9.65, p <0.0005).
This indicates that the patter n of increas e in difficulty of discerning the target caused by an increase in nu m be r of strea m s was differen t for the two mem o ry conditio n s. This was investigate d by more detailed analyses: Bonferroni t- test s were used to deter mi ne the points at which there were significan t increases in incorrect respo n s e s.
In the Echoic conditio n there was a significan t increase in the amou n t of incorrect respo n s e s when the numbe r of simulta ne o u s strea m s change d from 3 to 4. In the Non - echoic condition there was a significant increas e in incorrect respo n s e s when the numbe r of
simulta n e o u s stream s increase d from 2 to 3. This indicates that when echoic memo ry is free to be used, there is a proces sing capacity of 3 audito ry strea m s, yet when echoic memory usage is supp re s s e d there is a capacity of 2 audito ry strea m s (See Figure 2).
Figure 2 - Percent age of errors as a function of echoic mem ory use and num b e r of simulta n e o u s auditory stream s
0 10 20 30 40 50 60 70 80
1 2 3 4
Number of Streams Playing Simultaneously
Mean Percentage of Errors
Echoic Non-echoic
2.4 Analysis of Reaction Times
Mean target detection times as a function of echoic memory usage and the num b e r of simulta n e o u s strea m s are shown in Figure 3.
Figure 3 – Mean target detection times (in secon d s) as a function of echoic mem ory usage and the nu m b er of simulta ne o u s auditory
strea m s
A 2- way repeate d meas u r es ANOVA was perfor m e d on these data.
Echoic memory led to significantly quicker respo ns e rates to the target stimuli (F 17, 1 = 41.5, p < 0.0005). The num be r of strea m s prese n t e d simulta neo u sly had a significant effect upon reaction times (F 29, 1 = 387.3, p <0.00 0 5). As the nu mbe r of strea m s played simulta n e o u s ly increase d, the time taken to respo n d to the target stimulu s significantly increase d. To check that the increase in reaction times was a linear increase, an unplann e d linear contra s t was run on the data. In both the echoic and non - echoic conditio ns there was a significant linear tren d. In the echoic condition F 7, 18 = 105.285, p < 0.00005. In the non - echoic conditio n F 7, 18 = 62.079, p
< 0.00005.
3 DISCUSSION
The findings of this study clearly indicate that echoic memo ry can play a role in models of auditory atten tio n. Attentional capacity was 3 auditory stream s when echoic memo ry was free to contrib ut e to
0 1 2 3 4 5
1 2 3 4
Number of Streams Playing Simultaneously
Response Time (secs)
Echoic Non-echoic
perfor m a n ce, but when echoic memo ry input was eliminat ed by articulatory supp re s sio n, the apparen t attentional proces sing capacity dro pp e d from 3 to 2 strea m s. Both these estimat e s have parallels in previou s work. The prese nt findings may help to explain discrep a n cies in earlier researc h, where echoic mem o ry may have contrib u t e d to capacity estimat e s in different degrees, depen di n g on the precise experi me n t al proced u r e s. For exam ple, certain research (Broadben t, 1958; Treism a n, 1964) which utilised shad owing techniq u es between 2 auditory sources, and which as such, did not prom o t e inpu t from echoic mem ory yielded the Gestaltian figure / g r o u n d theories of attentio n. However, more recent estima te s of attentio nal limitatio ns (Huron, 1989; Gregory, 1990; Ludlam, 1993) whose experime n t al techniq ue s involved target detectio n in multiple strea m s (a proced u re in which echoic mem ory can contrib u t e to the perfor m a n ce) prod uce d the higher attentio nal limit estima t e of 3 strea m s.
Within the general framewo rk of models of attention, the prese n t finding s suggest a more flexible mode and complex model of attentio n al processing. The finding of 3 - stream capacity before a sudd e n increase in errors indicates a bimo dal syste m, an elemen t of parallel proces sing giving way under increase d press u r e to a more serial mode. But the fact that this capacity is lower under the Non - echoic conditio n indicates that the appare n t parallel proces sing is depen d e n t on sup po r t from echoic memo ry. Further m o r e, the fact that reaction times show a steady increase propo r tio nal to the num be r of simulta n eo u s strea m s implies an underlying serial proces s, even under condition s of echoic mem ory contrib u tio n.
However, under natural listening condition s echoic memory may be viewed as an integral part of the atten tio n al syste m. For this reaso n the mecha ni s m of atten tion might best be depicted as a predo mi n a n tly parallel and elastic proces si ng syste m showing evidence of strain (increasing reaction times) but partially absorbing increase d dema n d s until its limit is reached. At this point proces sing changes to a pred o mi n a n tly serial mode (non - linearity of errors), the elastic limit being influence d by specific situation al characteris tics.
The demo n s t r a ti on of the effect of mem o ry on attention process e s shows that the two concep t s are virtually in a symbiotic relation s hi p. Previous researc h has failed to acknowledge this. It appear s that the Wundtian ques t for the atom s of perception has
reached as far as it can into purely auditory attention. It seems impos sible to only meas ur e atten tio n witho u t involving memory.
4 REFERENCES
Broadbe nt, D. E. (1958). Perceptio n And Com m u n ication . Oxford : Perga mo n.
Treism a n, A. M. (1964). Selective Attention In Man. British Medical Bulletin . 20, 12 - 16.
Sloboda, J. A. and Edworth y, J. (1981). Attending To Two Melodies At Once: The Effects Of Key Related ness. Psychology Of Music. 9 (1), 39 - 43.
Huron, D. (1989). Voice Denu m e r a bility In Polyphonic Music Of Homogeno us Timbres. Music Percep tio n. 6 (4), 361 - 382.
Gregory, A.H. (1990). Listening To Polypho nic Music. Psychology Of Music. 18 (2), 163 - 170.
Ludla m, R. (1993). Undergr a du a t e Dissertation, City University, London.
Cowan, N. (1984). On Short And Long Auditory Stores. Psychological Bulletin . 96 (2), 341 - 370.
Baddeley, A. D. and Hitch, G. J. (1974). Working Memory. In G.H.
Bower (Ed.) The Psychology Of Learning And Motivation.
(Vol. 8). London: Acade mic Press.
Baddeley, A. D., Grant, S., Wight, E. and Tho m pson, N. (1975).
Imagery and Visual Working Memory. In Rabbitt, P. M. A. and Dornic, S. (Eds.) Attentio n And Perfor m a n ce (Vol.5). London:
Acade mic Press.
Sala mé, P. and Baddeley, A. (1982). Disruption Of The Short - Ter m Memor y By Unattende d Speech. Implications For The Structure Of Working Memor y. Journal Of Verbal And Learning Behaviour. 21 (2), 150 - 164.
Keller, T. A., Cowa n, N., and Saults, J. S. (1995). Can Memory For Tone Pitch Be Rehearsed? Journal Of Experiment al Psychology:
Learning, Memory and Cognition. 21 (3), 635 - 645.
Bregm a n , A. S., (1990). Auditory Scene Analysis: The Percept u al Organisa tio n Of Sound. MIT Press. Cam brid ge.
