Domagoj Švegar & Dražen Domijan dsvegar@ffri.hr ddomijan@ffri.hr

1. 8th Alps-Adria Psychology Conference October 2-4, 2008, Ljubljana, Slovenia A Comparison of Methods for Estimating the Capacity of Visual Working Memory: Examination of Encoding Limitations Domagoj Švegar & Dražen Domijan dsvegar@ffri.hrddomijan@ffri.hr University of Rijeka, Rijeka, Croatia Faculty of Arts and Sciences Department of Psychology

2. Introduction (Methodology of visual working memory capacity research) • Change detection paradigm (Phillips, 1974) • stage 1: initial presentation • stage 2: retention interval • stage 3: test • Quantitative approach in estimating visual working memory capacity (Pashler, 1988) Alps-Adria 2008

3. Change detection paradigm(stage 1: initial presentation) Alps-Adria 2008

4. Change detection paradigm(stage 2: retention interval) Alps-Adria 2008

5. Change detection paradigm(stage 3: test) Alps-Adria 2008

6. Pashler’s quantitative approach • C– number of objects stored in memory(visual working memory capacity) • H– is hit rate • IP– total number of items presented in a display • FA – false alarm rate Alps-Adria 2008

7. The main problem of the present study • All experiments that investigate visual working memory capacity are based on the same change-detection paradigm. In these experiments, objects are simultaneously displayed before retention interval. Displays usually contain up to 16 objects, and their duration is extremely short: usually initial exposition lasts 100-150 ms. • Visual objects can not be adequately encoded if attention is not directed to them(e.g. Coltheart, 1980; Irwin, 1991; Resnink, O'Regan, & Clark, 1997). • Attention can simultaneously be directed to not more then four visual objects(e.g. Pylyshyn & Storm, 1988; Yantis, 1992). • Not more then four objects can be stored into visual working memory(e.g. Vogel et al., 2001; Wheeler & Treisman, 2002; Alvarez & Cavanagh, 2004). Therefore: Is visual working memory capacity indeed soextremely limited, or it is erroneously underestimated due to inadequate encoding of stimuli? Alps-Adria 2008

8. The main problem of the present study • The central idea of the present study was to force participants to attend to every single displayed stimulus. If that could be accomplished, then it would be certain that all of the stimuli were successfully encoded into memory. Directing attention to each presented stimulus was ensured by successive initial stimuli presentation. Participants' performance in that condition was compared to their performance in the classical change detection paradigm with simultaneous initial presentation of stimuli. • Besides the initial presentation, testing of memory was also varied. In one condition memory was tested with partial test-displays, and in the other, full test-displays were applied. Alps-Adria 2008

9. Method • 2 x 2 experimental design was used, with type of initial presentation (simultaneous / successive) and type of test situation (partial test-display / full test-display) as factors. Visual working memory capacities were estimated and compared for each of these four conditions.Each participant went through both display conditions, whiletype of initial presentation was varied between subjects. • Thirty-seven undergraduate psychology students (age range 19-28, 2 male) from The University of Rijeka, Croatia, participated in the experiment. Alps-Adria 2008

10. Simultaneous presentation of stimuli Alps-Adria 2008

11. Successive presentation of stimuli Alps-Adria 2008

12. Method • All relevant factors besides the type of initial presentation and the type of test display were held constant: 1. set size was not varied – all trials consisted of 8 squares 2. total interval of initial presentation was always 500 ms 3. colors and locations of stimuli were selected randomly for each trial, but in order to assure that tasks are equaly demanding in all conditions, the same 60 trials were used in every experimental condition (only their order of presentation was randomly varied across participants) 4. verbal load procedure was identical in all conditions Alps-Adria 2008

13. Results and discussion • Participants’ performance was measured via: 1. Percentage of correct answers 2. Pashler’s capacity estimates 3. Reaction times (this was a supplementary measure) • All these dependent variables were subjected to seperate mixed two-way analyses of variance with the type of initial presentation as a between subject factor, and the type of test display as a within subject factor. Alps-Adria 2008

14. Effects of the type of initial presentation • All analyses have shown that the main effect of the type of initial presentation was not significant. • The interaction between the type of initial presentation and the type of test display was also not significant. • These findings do not corroborate the assumption according to which visual working memory capacity estimates, obtained in previous studies, were underrated due to encoding limitations. Alps-Adria 2008

15. Effects of the type of test display • The main effect of the type of test display was unclear. • When measured via percentage of correct answers or via reaction time, it was not significant. • However, when Pashler’s estimates were entered into ANOVA as a dependent measure, the effect of test display type was significant: visual working memory capacity was higher in partial test display condition, compared to full test display condition. Alps-Adria 2008

16. Effects of the type of test display • Why did these conflicting results occur? • Several simulations have shown that Pashler’s procedure: • has a tendency to overestimate true visual working memory capacity • is dependent on the criterion of the responding. • Analyses of hit and false alarm rates have shown that true memory capacity did not differ as a function of the type of test display – instead of that, differences found by Pashler’s formula were caused by differences in participants’ criterion of responding. Alps-Adria 2008

17. New method for assessing visual working memory capacity • Since Pashler’s procedure tends to overestimate true visual working memory capacity, is heavily dependent on the criterion of the responding and sometimes gives senseless estimates, a better procedure of visual working memory assessment was constructed. • The new method is based on the percentage of correct answers, rather then on hit and false alarm rates: where: C% is visual working memory capacity estimated via percentage of correctanswers PC is percentage of correct answers IP is set size (number of initially presented items). Alps-Adria 2008

18. Conclusion • Results of the present study are not consistent with the hypothesis that visual working memory capacity is underestimated due to inadequate encoding of stimuli. • New formula of visual working memory capacity estimation is constructed: Alps-Adria 2008