The effect of cognitive interference tasks on visuospatial short-term memory
[This snappily-titled piece is my thesis from 1997, and a masterful work of psychobabbling essay-speak it is too. I don't expect anyone to actually read the thing but I'm too busy at work at the moment to write any reviews, as everyone scrambles to get their documents out by the end of 2006, so I'm just updating my blog in cheap and easy ways by copying stuff I'd written earlier. Enjoy. Or, more probably, don't.
In recent experiments regarding the nature of the visuospatial sketchpad (VSSP), there has been much unresolved debate as to whether encoding is primarily visual or spatial (eg. Toms et al, 1994) and the extent to which the central executive is involved in the short-term storage of visuospatial information (eg. Morris, 1987). Using the dot-pair comparison paradigm suggested by Hole (1996), this paper reports an experiment designed to test the relative interference effects of a series of noise-screens and noise judgement tasks on the maintenance of data in the VSSP. The results of Hole (1996)’s unattended noise-dot condition were replicated, but other forms of interference employing visual processing or the central executive had no significant effect on performance. This suggests that the central executive is unnecessary for maintenance of spatial information in the VSSP, and that visual noise alone is not sufficient to disrupt it.
The notion of a modular “working memory” is one which has come under much investigation in the past few decades. Among its many proponents are researchers such as Baddely, who have pointed out the way in which our concepts of memory are fragmenting as knowledge of its properties grows, even as far back as the initial distinction between long-term and short-term memory, which few would now dispute (Baddely, 1986). The concept has evolved through models of modality (after Broadbent, 1958) and Levels of Processing (eg. Craik and Lockhart, 1972, cited in Baddely, 1986); the most commonly-held view now is in accordance with Baddely (1986)’s framework of a modular working memory of at least 3 subsystems. These consist of an “articulatory loop” for the encoding and maintenance of verbal speech-based material, a visuospatial sketchpad or scratchpad - Baddely himself prefers the term “sketchpad”, as “scratchpad” carries equal and misleading implications of verbal notes alongside visual or spatial data (Baddely, 1986, p109) - and a central executive function which serves to co-ordinate or control the two slave systems.
It is the articulatory loop which has attracted the bulk of the research in this area, whether for ease of experimentation (Logie, 1986) or otherwise, and has produced some interesting results which seem to support Baddely’s working-memory model. For example, the technique of “articulatory suppression” has enabled researchers to demonstrate the relative independence of the articulatory loop from the central executive, in that it selectively interferes with tasks associated principally with the articulatory loop, such as digit span, without affecting verbal reasoning and other central executive functions (Hitch and Baddely, 1976). More recently, however, there has been a growing body if interest in the previously neglected area of the VSSP.
One of the main problems with investigations in this area is the lack of an obvious equivalent to articulatory suppression (Logie, 1986); being confined to the articulatory loop, this technique has little or no effect on maintenance of the visuospatial short-term store. What was needed was some form of visual or spatial distraction to produce a similar effect (Hitch, 1984). Tracking tasks provided a useful tool for creation of visuospatial interference, motor movements having a direct effect on spatial processing (Quinn, 1976) and they have had the additional benefit of serving to identify a distinction between the visual and spatial elements of the VSSP when non-visual tracking was used (Baddely and Lieberman, 1980). However, to term this an accurate equivalent to articulatory suppression would be erroneous, as the latter is an unskilled activity subject to far fewer effects of fatigue or practice (Logie, 1986) and as such probably has less involvement with the central executive. Attempting to filter out the effects of the central executive has been a common problem in experiments on the VSSP, as it is thought to be an essential feature of both encoding and retrieval (Morris, 1987). The extent to which any interference is specific to either of the slave systems is therefore difficult to determine; however, experiments by Morris (1987) and Toms et al (1994) have shown that during maintenance, the VSSP is passive and thus remains unaffected by any general load of the central executive not directly relating to visual or spatial material. This runs contrary to the findings of Phillips and Christie (1977) and Dale (1973), who suggested that the short-term store relied on general purpose resources and was subject to nonspecific forms of distraction.
If it is therefore necessary to use visuospatial stimuli to interfere with the maintenance of information in the VSSP, could this then include all forms of visuospatial distraction? If Jones et al (1995) are correct in their assertion that short-term stores of verbal and visuospatial data are functionally equivalent, then it would be counterintuitive to believe so, as it has been shown that only very specific forms of auditory stimulus are able to interfere with verbal speech-based short-term memory (Salame & Baddely, 1989). It seems logical to assume that the stimuli needed to disrupt the passive activity must necessarily be similarly specific. It is the exact nature of these interference stimuli which provide more detailed information on the workings of the VSSP and its potential internal subsystems.
The obvious distinction to make would be that between visual and spatial data. It has already been noted that such a distinction exists, as was demonstrated by Baddely and Lieberman (1980) in a tracking task using auditory feedback with no visual input whatsoever. This was disputed by Beech (1984), whose broadly similar experiment showed interference from both visual and spatial quarters; however, it has been suggested that these conflicting results were due to the spatial nature of the response requirements (Toms et al, 1994). Either way, the relationship between visual and spatial processing is one which has considerable importance for the understanding of the VSSP.
One of the paradigms most often used for investigation of this area is the Brooks Spatial Matrix task (Brooks, 1967) which involves mental imagery. The equivalence of mental imagery to actual visuospatial input has been demonstrated by many previous studies (eg. Brooks, 1967, Quinn, 1976, Logie, 1986) with use of imagined matrices and visual mnemonics, and their abilities to selectively interfere with real data. The Brooks task, where subjects are asked to imagine the locations of digits on a 4x4 grid, was generally understood to be primarily spatial in nature and thus treated as a good basic spatial test, but more recent findings by Toms et al (1994) have shown effects of visual interference on maintenance of this grid. This has even led the authors to suggest that there may be no spatial store as such, merely a “spatial processing device, perhaps ensconced within the central executive complex” which in certain circumstances acts upon the visual store (Toms et al, 1994, p142). However, given that this conflicts with the theories of many other researchers and new evidence from neuropsychology, it is likely that this is due to some variance in experimental procedure.
A theory which might go some way towards explaining this discrepancy is that of a difference between encoding of patterns, pictures with possible additional semantic coding and simpler spatial information (Logie, 1986, Hole, 1996). It seems to be clear that a all forms of visual material have obligatory access to the VSSP; the amount of disturbance caused to the information stored there passively depends largely on the similarity between the interference and stored data. Much research has shown that the relative complexity of the interference stimuli has a marked effect on the extent of the distraction produced (eg. Logie, 1986, Magnussen, 1988); it is a distinct possibility that “the mechanisms of underlying memory for small-scale spatial relationships might be more susceptible to disruption than those responsible for large-scale relationships” (Hole, 1996, p54).
A problem with many experiments in the past is the lack of any obvious attempts to choose primary or secondary stimuli on the basis of anything other than their visual or spatial components. Logie, in his series of 1986 experiments, tried to rectify this to a certain extent, taking into account both visual complexity and semantic content, but the secondary tasks still do not appear to have been matched for complexity with the primary ones, a visual imagery mnemonic and rote rehearsal strategy. The use of verbal material in any case leaves the investigation open to possible effects of verbal recoding during maintenance (Morris, 1987); a paradigm was needed which as far as possible took note of all these complicating factors.
The experiment on which this paper is based, that by Hole (1996), seems to do just that. The primary task of spatial interval estimation is sufficiently taxing without being overcomplex, and the verbal element is minimised; both of these factors preclude excessive involvement of the central executive and thus produce a “cleaner” result. It is also easier to suggest equivalence of complexity between, for example, the dot-pair stimulus and a brightness discrimination as both are relatively simple, unlike in cases such as Logie’s (1986) where the primary and secondary stimuli consisted of material varying from pegword imagery and line drawings to coloured screens and matrix patterns.
Research already conducted on this paradigm has shown that without interference, it is possible to maintain accurate spatial representations of the two dots across fairly long periods of time, but with a pair of interference dots between the two comparison pairs, thresholds were made dramatically worse regardless of whether the subject was obliged to attend to this interference (Hole, 1996). These results have provided more evidence of obligatory access of visual material to the VSSP, but as the interference was of the same visuospatial nature as the primary stimuli, it is unclear which aspect of this causes the spatial judgement to be affected. The experiment detailed here is an attempt to clear up some of this confusion.
Morris (1987) suggests that the central executive is required for encoding and retrieval of all data in the VSSP; as all visual data has privileged access to the VSSP regardless of subjects’ intentions, the central executive must necessarily be activated for encoding of the interference stimuli. It is therefore conceivable that retrieval failure is merely due to an overload at the central executive level. To test for this effect, two of the conditions in this experiment require subjects to make a comparison judgement on the noise stimuli, an activity which requires central processing. One of these tasks is of brightness discrimination, which has a strong visual component without involving spatial factors; this provides a secondary visual task of comparative complexity to the primary visuospatial one, and should measure the effect of the central executive combined with visual noise. The other noise task is of temporal discrimination, with no spatial components and visual ones which can be matched against those of one of the controls (condition 4 - see below). The results of this experiment should show which of these factors has the greatest effect on maintenance of the visuospatial data stored in the VSSP.
10 subjects aged between 19 and 40 took part in this experiment; 6 were male and 4 were female. Three of the subjects wore glasses; the rest had either normal vision or contact lenses which would not have caused any potential interference from reflection. All of the subjects were informed of the purpose of the experiment, but only four had any knowledge of psychology and thus the experiment’s implications; only one subject had previous experience of psychometric tests.
The experiment was conducted in a room with both windows and normal fluorescent lighting, though care was taken to ensure minimum light-reflection on the screens. The screens in question were of a VGA colour monitor attached to an IBM PS/2 computer, through which the stimuli were presented. The subjects were seated at an approximate distance of 60cm from the screen, and responses were entered through the computer keyboard. Because of the communal nature of the room used, there was inevitably some auditory distraction throughout much of the experiment.
c) Design and Procedure
The experiment consisted of 5 conditions performed by each subject in a random order. Each condition involved 60 trials, the trials being of equal length across all conditions. The instructions were provided verbally by the experimenter at the start of each condition but the experiment itself was controlled by a computer (see above). Each trial in all conditions was made up of four separate stimulus screens, each lasting 2 seconds, with the ISI being 1s. For purposes of description, the stimulus screens shall be labelled a, a1, b1 and b; they were always presented in this order.
In all the conditions, screens a and b provided the primary spatial-judgement task. As in the Hole (1996) experiment, this was achieved by a two-alternative forced-choice procedure, with the subject having to decide which of two horizontal pairs of dots was spaced the wider, one pair appearing on screen a and the other on screen b. The dots used were 3mm square with an approximate luminance of 300cd m¯², against a back background of approximate luminance 105cd m¯².
From a starting point of 65mm (measured from the centres of the dots), the distance between the standard stimulus dots was randomly varied by 10% in either direction to prevent the subject merely encoding that standard as their basis for comparison. Each comparison stimulus distance was calculated in much the same way, but then had up to 40% of that length either added to or subtracted from it. An adaptive program was used to calculate thresholds (using Weber fractions) based on subjects’ spatial judgements; decisions were registered by pressing one of two keys. The pairs of dots appeared in approximately the centre of the screen, with random variation being introduced horizontally and vertically on each trial to prevent subjects’ use of a screen-related strategy.
Screens a1 and b1 varied between conditions. Condition 1 replicated the un-attended noise-dot condition from Hole (1996) in that screen a1 contained a pair of irrelevant dots, screen b1 being blank. The subject were obliged to look at the screen during the presentation of all noise stimuli, but were aware that these dots were merely interference. No secondary task was present in this condition.
Condition 2, however, contained a secondary brightness-discrimination task in screens a1 and b1. Two grey screens of varying brightness were presented in each trial, the ISI screens again being black. In condition 3 the secondary task was of temporal discrimination; the grey screens were of identical luminance (this was varied randomly between trials) but each was on view for a slightly different length of time. At the end of each trial in these conditions, subjects were required to state which screen had been respectively brighter or longer-lasting; these judgements were made in a second two-alternative forced-choice procedure which occurred after the primary dot-task decision for each trial had been registered.
Conditions 4 and 5 were the controls. Subjects were still obliged to look at the screen but no secondary judgements were required. In condition 4, the screens a1 and b1 were an identical shade of grey (again, this was varied between trials) and of identical duration to measure the effect of unattended grey-screen interference. The last condition was merely to measure the effects of the time-delay without noise, as both a1 and b1 consisted of unchanging black screens.
The subjects received no feedback of their performance during each condition other than what could be gathered from the workings of the adaptive program - if they were doing badly it would become apparent from the increased easiness of the task. At the beginning of each trial, subjects were informed of how many trials they had already completed, which may have had some bearing on a possible “boredom effect”; with each condition lasting between 15 and 20 minutes depending on subjects’ response time, the whole experiment could last up to 1 hour 40 minutes per subject. However, subjects were not encouraged to hurry or treat this as a response-time task; in fact, they could pause at any point during the conditions if they so chose. Most subjects chose to complete the experiment in more than one session, though some managed it in only one sitting. The experimenter was present at all times to ensure the subjects were looking at the screens during all the relevant noise conditions.
Two of the subjects’ results were discounted following their extremely poor performance on the control task (condition 5); it was judged that their lack of concentration or skill even at this level rendered their data useless to this investigation.
Figure 1 Mean thresholds for each condition
|Condition 1||Condition 2||Condition 3||Condition 4||Condition 5|
From the means (Figure 1), it can be seen that the main effect is definitely that of the interference dots (condition 1), with a mean threshold of over twice the size of either of the controls (conditions 4 and 5). The unattended grey-screens (condition 4) had very little effect on thresholds, and the results of the two noise-task judgement conditions are very similar to each other.
A one-way repeated-measures analysis of variance carried out on the thresholds proved to be significant [F(4, 35) = 2.8, p<0.05]. A series of t-tests revealed the only significant differences to be between condition 1 and the two controls (C1-C4, t=3.13. p<0.05; C1-C5, t=2.92, p<0.05). Interestingly, the difference between the two noise-task judgement conditions is highly insignificant (t=0.35, p<0.05), though not as insignificant as that between the two controls (t=0.19, p<0.05). No correlation was found between performance on the primary and secondary tasks.
As expected, the results of this experiment have largely supported the findings of Hole (1996), with regard to the visuospatial noise having obligatory access to the VSSP and causing interference to the information stored there. Condition 1, being very similar to one of the conditions in Hole’s Experiment 2 (1996) but with a slight temporal variation, produced much the same results as the latter; this is consistent with Hole’s claim that his results are “highly replicable” (Hole, 1996, p62). The temporal difference between the two experiments, caused by the extra stimulus screen in the present one, is unlikely to have had any effect, as Hole’s Experiment 1 (1996) showed that spatial information of this complexity in the short-term store decayed only gradually (see Dale, 1973 for the decay effects on more complex visuospatial information). The main difference in results is the much higher thresholds found here; none of Hole’s subjects obtained thresholds higher than 14% of the spatial interval in any condition, whereas in this experiment some thresholds ran as high as 64.54%.
The only significant effect was that of the visuospatial interference, the noise-dots of condition 1. This means that the effects of the other noise conditions can be judged irrelevant to maintenance of the visuospatial data presented on screen a. The fact that the brightness-discrimination had no significant effect would seem to run contrary to Toms’ assertion that all visual input has a distracting effect regardless of complexity (Toms et al, 1994). It may be the case, however, that the spatial component of this spatial-interval judgement task has such priority over the visual component that no non-spatial distraction is sufficient to disrupt it.
That neither of the secondary-task conditions had any significant effect supports Morris’s view that, unless interfered with directly by relevant visuospatial input, the VSSP is a passive store unaffected by the workings of the central executive (Morris, 1987). In these conditions, the central executive must necessarily have been active, but this has not proved to affect maintenance of the data stored; the fact that the minimal interference caused by both tasks is so similar indicates that the only effect occurring here was general distraction of a nonspecific nature, and thus unable to selectively distort the contents of the spatial store.
It would have been useful here to have included a sixth condition featuring a spatial non-visual judgement task, which would have helped to resolve the matter of whether this information is stored in a system specific to visuospatial data or a pan-sensory spatial one, but this was not possible for reasons of time, resources and the patience of the subjects. The feasibility of such further experimentation, however, is called into question by the problems of producing a purely spatial task, as unlike visual processing, a spatial system is not necessarily tied to any particular sense. It may be the case that some spatial store is common to all spatial input, but at present, research has been largely unable to filter out all irrelevant data, for example motor action from kinaesthetic spatial coding (Quinn, 1976, Smyth & Pelky, 1992); it would also be difficult to draw parallels between results obtained from such different sources.
Another point that should be addresses here is one mentioned earlier, the possibility of verbal recoding during maintenance (Morris, 1987). While it would be extremely difficult to do so effectively in the case of this experiment, the phenomenon should not be discounted. It seems unlikely that subjects would not try to use all possible strategies at their disposal, and a verbal one would be almost inevitable. This would certainly have some bearing on the results obtained, given the probable involuntary coding of the noise dots alongside the comparison stimulus in condition 1; this coding is a lot more likely to become a distraction than the different codes of “brighter” or “darker” in condition 2, or temporal adjectives in condition 3. Perhaps in future experimentation, articulatory suppression could be used to lessen the possibility of this happening.
Despite the growing body of evidence in support of this model of working memory, it would be unwise to jump to any conclusions. The objections exist and need to be answered if a truly accurate picture of the subsystems of working memory is to be created. However, these results do suggest that research is heading along the right lines; it would be difficult now to dispute the concepts of obligatory access and visuospatial interference, even when taking the possibility of verbal recoding or other strategy into account. The two-dot spatial-interval judgement task seems to be a useful paradigm for research into the nature of the VSSP, and the experiment presented here provides some new information on its possible workings. Further study on the matter of spatial coding may prove to be the direction in which the most useful insights lie.
- Baddely, A.D. (1986) Working Memory Oxford: Oxford University Press
- Baddely, A.D. & Lieberman, K. (1980) Spatial Working Memory In R. Nickerson (Ed.), Attention and Performance VIII pp521-539 Hillsdale, NJ: Lawrence Erlbaum Associates
- Beech, J. (1984) The effects of visual and spatial interference on spatial working memory Journal of General Psychology 110, 141-149
- Broadbent, D.E. (1958) Perception and Communication London: Pergamon Press
- Dale, H.C.A. (1973) Short-term memory for visual information British Journal of Psychology 64, 1-8
- Hitch, G.J. (1984) Working Memory Psychological Medicine 14, 265-271
- Hitch, G.J. & Baddely, A.D. (1976) Verbal reasoning and working memory Quarterly Journal of Experimental Psychology 29, 603-621
- Hole, G.J. (1996) Decay and interference effects in visuospatial short-term memory Perception 25, 53-64
- Jones, D., Farrand, P., Stuart, G. & Morris, N. (1995) Functional equivalence of verbal and spatial information in serial short-term memory Journal of Experimental Psychology 21(4), 1008-1018
- Logie, R.H. (1986) Visuo-spatial processing in working memory Quarterly Journal of Experimental Psychology 38A, 229-247
- Magnussen, S., Asplund, R., Dyrnes, S., & Greenlee, M.W. (1988) Spatial selectivity of the short-term memory image Perception 17(3), A54-A55
- Morris, N. (1987) Exploring the visuo-spatial scratchpad Quarterly Journal of Experimental Psychology 39A, 409-430
- Phillips, W.A. & Christie, D.F.M. (1977) Interference with visualisation Quarterly Journal of Experimental Psychology 29, 637-650
- Quinn, J.G. (1976) Imagery and the representation of spatial information Unpublished Thesis
- Salame, P. & Baddely, A.D. (1989) Effects of background music on phonological short- term memory Quarterly Journal of Experimental Psychology 41A, 107-122
- Smyth, M.M. & Pelky, P.L. (1992) Short-term retention of spatial information British Journal of Psychology 83, 359-374
- Toms, M., Morris, N., & Foley, P. (1994) Characteristics of visual interference with visuospatial working memory British Journal of Psychology 85, 131-144
Brooks, L.R. (1967) The suppression of visualisation by reading Quarterly Journal of Experimental Psychology 19(4), 289-299