Top-down influences on the time-course of visual attention: When the attentional blink meets word superiority effect Maria V. Falikman Lomonosov Moscow University Research supported by the Russian Fund for Basic Researches, grant # 03-06-80191 Where cognitive science was born… “… I went away from the symposium with the strong conviction, more intuitive than rational, that human experimental psychology, theoretical linguistics, and computer simulation of cognitive processes were all part of a larger whole...” (George Miller on the 1956 MIT Symposium) Types of processing (information flow) Task, Experience Top-down (conceptually-driven) Conscious percept Bottom-up (data-driven) Stimulus Top-down influences on visual attention Unconscious (implicit) Conscious (explicit) Context-based or memory-based Goal-directed or task-dependent Contextual cueing Priming (+/-) Schemata ... Strategic regulation of perceptual task accomplishment Rapid Serial Visual Presentation (RSVP) RA PI O U R N V P S E A T L I DS E RI AL The Attentional Blink (AB) under RSVP conditions RA PI DS AL The Attentional Blink (AB) under RSVP conditions RA PI DS ER IA L The Attentional Blink (AB) under RSVP conditions RA PI DS AL The Attentional Blink (AB) under RSVP conditions RA PI DS ER IA L The Attentional Blink (AB) under RSVP conditions RA PI DS AL The Attentional Blink (AB) under RSVP conditions RA PI DS ER IA L The Attentional Blink (AB) under RSVP conditions RA PI DS AL The Attentional Blink (AB) under RSVP conditions RA PI DS ER IA L The Attentional Blink (AB) under RSVP conditions RA PI DS AL The Attentional Blink (AB) under RSVP conditions RA PI DS ER IA L The AB effect replicated: Results of the standard dual-task RSVP experiment (2000) 100 Experimental condition % Correct Reports 80 Control condition 60 40 20 0 0 1 2 3 4 5 Probe relative serial position 6 Why a blink? Blindness! Hybrid models Perceptual Amnesia! system Object reconfiguration substitution Central theory interference theory Two-stage model Attentional dwell-time Interference theory Attentional model gate model Models of selective attention and models of its blink Early selection / filter Attentional gate (Broadbent, 1958) (Raymond et al, 1992) Late selection Interference model (Deutsch & Deutsch, 1963) (Shapiro et al, 1994) Pertinence model Two-stage model (Norman, 1968) (Chun & Potter, 1995) Multiple selection “Smart selection” (Johnston & Heinz, 1979) (Shapiro & Luck, 1999) Flexible selection (Yantis & Johnston, 1990) “Intelligent filtering system”(Enns et al, 2001) Perceptual cycle Reentrant model (Neisser, 1976) (Di Lollo et al, 2000) Quo vadis? Attention for action D.A. Allport O. Neumann A.H.C. van der Heijden Attention as an aspect of action Gippenreiter, 1983 [rus] 1986 or a set of “task-related mechanisms” (e.g. Deacon & Shelley-Tremblay, 2000) The prototypical “bottleneck” model T1 Early stage T2 LC-stage Early stage T1 report ... LC-stage T2 The AB effect: my old results again 100 Experimental condition % Correct Reports 80 Control condition 60 40 20 0 0 1 2 3 4 5 Probe relative serial position 6 The AB effect might be due to the inability of perceptual system to obtain further information until the previous goal-directed act (conscious target identification) is completed. Then, the AB could be modulated through the change of the size of the observer’s activity “units”: if we incorporate the 1st target into a larger perceptual “object”, the AB will probably not interrupt perception until the end of the act. “The difficulty is not to combine stimuli, but rather to deal with them independently at the same time... capacity limits occur where stimuli have to be kept apart, not where they have to be combined…” (Neumann, 1987, p.363) “Attentional units” larger than letters? W O R D S Stimuli examples: Russian letters PRINTED WRITTEN The word superiority effect Better performance on letters within a word than on random letters: • faster recognition • more letters perceived within a brief (10 ms) presentation “I find it takes about twice as long to read (aloud, as fast as possible) words which have no connexion as words which make sentences, and letters which have no connexions as letters which make words. When the words make sentences and the letters words, not only do the processes of seeing and naming overlap, but by one mental effort the subject can recognize a whole group of words or letters, and by one will-act choose the motions to be made in naming them, so that the rate at which the words and letters are read is really only limited by the maximum rapidity at which the speech organs can be moved...” (Cattell, 1886, p.64). The word superiority effect Further explorations: Reicher, 1969 Wheeler, 1970 McClelland, 1976 ... The word superiority effect R The word superiority effect WOR D The word superiority effect WOR D Some examples of “mutable” words (main experimental session) I ...S …W A P T R E I R T # E # L …F AI VL O RT # ...P G ...B …B RR II DD EG #E # Hypothesis IF the AB is due to a certain bottom-up controlled limitation (“a bottleneck”), it will leave its signature on the observers’ performance on mutable words. IF it is rather shaped by top-down processes in the information processing system, it will either disappear or be diminished in the word-reading task. Comparison of correct reports on separate letters in mutable words and on a probe X under “standard” AB conditions (2000) 100 Dual task (probe X) Single task (probe X) Mutable words % Correct Reports 80 60 40 Word identification point? 20 0 1 2 3 4 5 Probe relative serial position 6 There is a blink on mutable words! Individual differences. 100 Subject 6 % Correct Reports 80 Subject 17 60 Mean 40 20 0 1 2 3 4 Probe relative serial position 5 And what about the overall result? 100 % Correct Reports 80 Mentioning a position for all word strings 60 40 20 0 1 2 3 4 Probe relative serial position 5 The ACTUAL experimental design: Preceding control session: identify the type of a white letter and name three letters following it (after Weichselgartner & Sperling, 1987) Main experimental session: identify the type of a white letter and read a word beginning with this letter Concluding control session: identify the type of a white letter and name three letters following it (after Weichselgartner & Sperling, 1987) Strategic change: Results of framing sessions 50 Report probability, % 45 Preceeding session 40 35 Concluding session 30 25 20 15 10 5 0 1 2 3 4 5 Relative serial position 6 80 80 70 70 Report probability, % Report probability, % There could be no strategic changes! Individual differences again. 60 50 40 30 20 10 60 50 40 30 20 10 0 0 1 2 3 4 5 Relative serial position Subject 16 6 1 2 3 4 5 Relative serial position Subject 2 6 Control experiment: No practice effects 50 Preceeding session Report probability, % 45 40 35 Concluding session 30 25 20 15 No words noticed between! 10 5 0 1 2 3 4 5 Relative serial position 6 Word superiority in written vs. printed T1 task T1 type identification in the preceding control session: 89 3.9% T1 type identification in the wordreading session: 96.2 2.4 % Significant! (p<0.0001) T1 type identification in the concluding control session: 90.4 2.9% Significant! (p<0.0001) Subjective dual-task estimation through the experimental sessions (% of subjects) Preceding session Main session (wordreading) Two separate tasks Concluding session A single whole The AB seems to depend on the task and on subjective strategies. But strategies serve for achieving a goal, and the goal is to report on T1 etc. For that, verbal working memory consolidation or encoding is essential! Working memory consolidation: neurophysiological data P300 - a late ERP component, a correlate of working memory encoding and attentional load T2: suppressed for missed probes during the AB (Vogel et al, 1998; see also Rolke et al, 2001) T1: repeats the dynamics of the AB after T1 identification (McArthur et al, 1999) A possible model: two-stage compatible ... P ... P P H H T1 WM encoding... ... P R R Not yet! P A H H A R T1 + “word” WM encoding Report S E Time to start WM encoding: - I got the word! - My VSTM buffer is full! Conclusions Perceptual task accomplishment can in principle be organized in such a way, that, given a sufficient load of that limited-capacity block or process, the AB would either disappear or diminish and shift from its critical temporal interval. These changes can be determined both externally (by task conditions and requirements) and internally (by subjective strategy), that is characteristic for any human activity. Hurrah! Recent results working for this hypothesis: if T1 is an addition task (e.g. 2+3) and T2 is an answer (seems to be… 5!) there is no blink! (Kunar & Shapiro, in preparation) Conclusions Our results directly suggest that the attentional blink deficit does not represent merely the workings of a fixed capacity-limited mechanism for the perception of events occurring over time, but rather the structure of actions a subject has to perform in order to accomplish a given task. Missing controls? Pronounceable pseudowords - in progress Random strings, instruct to read words - in progress Words with ambiguous endings? - Thanks to Molly Potter for the brilliant idea! To be performed. More ideas? - Please keep them in mind until the discussion... Prospect: Visual search? Attentional switching (gating paradigm)? Metacontrast masking? Unconscious perceptual priming? Crowding? - YES (Fine, in press) Special thanks to Dr. Valeriy Romanov, Dr. Yuriy Dormashev (Moscow University) for inspiration of the AB research Ekaterina Pechenkova (Moscow University) for collaboration and invaluable discussions Our programmers: Suren Sagiyan George Kouryachy Thank you for your attention that hopefully didn’t blink (owing to the top-down regulation!)