Laboratory for Experimental Psychology KULeuven - Campus Kortrijk Eva Van den Bussche The mechanisms of subliminal semantic priming Promoter: Bert Reynvoet Research seminar 4th April 2007
Overview General background Goals Study 1: Masked priming effects in semantic categorization are independent of category size Study 2: The influence of attention on masked semantic priming effects General conclusions 2
Background Subliminal (or masked or unconscious) semantic priming ~> an unidentifiable visually masked stimulus (prime) facilitates the subsequent semantic classification of a visible related stimulus (target) Priming/congruency effect ~> improved responding (faster and/or more accurate) when primes are congruent to targets, relative to when prime-target pairs are incongruent 3
Background Example: is target smaller or larger than 5? Congruent trial Incongruent trial Prime 4 4 Target 2 8 Both < 5 Faster RTs Less errors One <, one > 5 Slower RTs More errors 4
Background Subliminal priming ~ long-standing debate in cognitive psychology Marcel (1983): unconsciously presented words prime semantic associates startling and counterintuitive Successful replications accumulated So did the skepticism (Holender, 1986) Last decade: methodological improvements more reliable research 5
Background New and improved research flow strong evidence for reproducible subliminal priming effects Nowadays: debate has shifted Subliminal priming exists! But: Are subliminal primes semantically processed? Which factors influence the emergence of subliminal semantic priming effects? 6
Goals Are subliminal primes semantically processed? Which stimuli/processes influence these masked semantic priming effects? Nature of stimuli Nature or primes Nature of task Category size Target set size Attention Shed light on these two debated issues! 7
Study 1 Study 1: Masked priming effects in semantic categorization are independent of category size Van den Bussche & Reynvoet, in press 8
Overview Theoretical background Predictions and goal Design Three experiments Conclusions 9
Are subliminal primes semantically processed? Three competing theories: Background 1. The semantic categorization hypothesis: subliminal primes are processed up to a semantic level (e.g. Dehaene et al., 1998) 10
Background 2. The direct motor specification hypothesis: subjects develop automatic S-R mappings, bypassing semantics: no priming for novel primes (Neumann & Klotz, 1994, e.g. Damian, 2001) 3. The category search model: subliminal priming only possible for small categories (e.g. Forster, 2004) 11
Predictions and Goal 1. The semantic categorization hypothesis: response congruency effects (RCEs), no matter what 2. The direct motor specification hypothesis: RCEs only for repeated primes and limited stimulus sets 3. The category search model: RCEs only for small categories Goal Study 1 = shed light on these different points of view 12
Design Congruent trial Incongruent trial time 17ms ###### LION time 17ms ###### LION 33ms horse 33ms mouse 480ms ###### 480ms ###### Smaller than dog Bigger than dog Smaller than dog Bigger than dog Fig. 1 Fig. 2 13
Design In all 3 experiments, targets never appeared as primes (novel primes) Response assignment was varied across subjects RTs and error rates were studied Prime visibility was always objectively measured ( apply same task to prime ): never differed from chance level!!! 14
Experiment 1 N 1 = 16 3 separate parts: 1a: categorize number words as smaller or larger than 5 (small category) 1b: categorize body parts as below or above the pelvis (small category) 1c: categorize animals as smaller or bigger than a dog (broad category) 15
Experiment 1: results 580 560 8ms, p <.05 RT (ms) 540 520 500 20ms, p <.001 25ms, p <.001 Congruent Incongruent 480 460 440 Experiment 1a Experiment 1b Experiment 1c Fig. 3 16
Experiment 1: conclusion Pattern analogous for analyses on error rates Conclusion: clear congruency effects for novel primes Damian (2001) for all categories Forster (2004) But: Our task Forster s task Small stimulus sets Experiments 2 and 3 17
Experiments 2 & 3 N 2 = 14 Exp. 2: categorize words as animals or objects N 3 = 16 Exp. 3: categorize words as animals or not Stimulus set size = 104 18
Experiments 2 & 3: results Experiment 2 Experiment 3 600 610 590 580 15ms, p <.001 600 590 11ms, p <.01 570 580 RT (ms) 560 Congruent Incongruent RT (ms) 570 560 Congruent Incongruent 550 550 540 540 530 530 520 Experiment 2: animal targets Experiment 2: object targets 520 Experiment 3: animal targets Experiment 3: non-animal targets Fig. 4 Fig. 5 Pattern analogous for analyses on error rates 19
Conclusions Study 1 Significant priming effects, even with novel primes, irrespective of stimulus set size, irrespective of category size and across different forms of tasks and task instructions! Evidence in favor of the semantic categorization hypothesis ~> subliminal primes are semantically processed! Based on these results: nature of primes, category size and target set size are not key influencing factors of subliminal semantic priming 20
Study 2 Study 2: The influence of attention on masked semantic priming effects Van den Bussche & Reynvoet, submitted 21
Overview Theoretical background Goals Four experiments Conclusions 22
Background Subliminal processing and attention are two of the most intensively studied areas in the field of semantic processing. Subliminal primes are thought to be semantically processed (cfr. Study 1) Attention is assumed to play an important role in semantic processing BUT: the role of attention in subliminal semantic processing remains unclear. 23
Background Two theoretical accounts: 1. Broadbent (1958): no identification is possible without attention (e.g. McCann et al., 1992) 2. Unconscious processing is automatic, independent of attention (e.g. Brown et al., 2002) However: research is scarce and contradictory It remains unclear whether attention is a prerequisite for subliminal semantic priming to occur. 24
Goals Goal = Clarify the influence of attention on subliminal semantic priming effects 1. Force subjects to divide attention between an increasing number of spatial locations does the strength of the RCEs decrease/disappear? (exp 1 and 2) 2. Manipulate whether the primes are attended or not does the strength of the RCEs decrease/disappear? (exp 3 and 4) 25
Experiments 1 and 2 N 1 = 24 Exp. 1: categorize numbers as smaller or larger than 5 N 2 = 24 Exp. 2: categorize words as animals or objects 26
Experiments 1 and 2 3 separate parts: a: stimuli were presented centrally on the screen b: stimuli could appear on two different locations on the screen c: stimuli could appear on three different locations on the screen Direction of attention was not manipulated: subjects had to divide their attention across all places 27
Experiments 1 and 2: design Congruent trial, where prime and target appear on different places for 1b time 480ms 33ms ## ## 17ms 2 ## ## ## ## 1 Smaller than 5 Larger than 5 Fig. 6 28
Experiments 1 and 2: design Incongruent trial, where prime and target appear on the same place for 2c 480ms time 33ms ##### ##### ##### 17ms ##### spin ##### ##### ##### ##### ##### ZAAG ##### Fig. 7 Animal Object 29
Experiments 1 and 2: design Novel and repeated primes Response assignment and order of the three parts were varied across subjects Prime visibility was always objectively measured ( apply same task to prime ): RCEs were independent of prime visibility! 30
Experiments 1 and 2: results Prime and target appeared on the same place *** p <.001 ** p <.01 * p <.05 p <.07 ns: not sign RCE (ms) 45 40 35 30 25 20 15 10 *** *** *** * *** *** * 5 0-5 -10 ns 1a 1b 1c 2a 2b 2c Experiment 1 Experiment 2 * ns ns Fig. 8 repeated novel 31
Experiments 1 and 2: results Prime and target appeared on different places *** p <.001 ** p <.01 * p <.05 p <.07 ns: not sign RCE (ms) 40 35 30 25 20 15 10 5 0-5 -10 ** *** *** ** *** ns ns ns 1b 1c 2b 2c Experiment 1 Experiment 2 Fig. 9 repeated novel 32
Experiment 1 and 2: conclusion Significant decrease of RCEs when attention had to be divided across more places on the screen ~ Broadbent s no identification without attention account. Effects for novel word primes are weak! 33
Goals Goal = Clarify the role of spatial attention on subliminal semantic priming effects 1. Force subjects to divide attention between an increasing number of spatial locations does the strength of the RCEs decrease/disappear? (exp 1 and 2) 2. Manipulate whether the primes are attended or not does the strength of the RCEs decrease/disappear? (exp 3 and 4) 34
Experiments 3 and 4 N 3 = 24 Categorize numbers as smaller or larger than 5 N 4 = 24 Exp. 2: categorize words as animals or objects 35
Experiments 3 and 4 2 separate parts: a: stimuli could appear on two different locations on the screen b: stimuli could appear on three different locations on the screen Direction of attention was manipulated! 36
Experiments 3 and 4 The target was 100% validly cued attention always drawn to place where target will appear Primes can appear on the same place as the target (= attended) or on a different place (= unattended) Do RCEs occur only when attention is paid to the prime? 37
Experiments 3 and 4: design Congruent trial, where the cue was not indicative for the prime in 3a 480ms time 67ms ## ## 33ms ## ++ 33ms ## ## 17ms 2 ## ## ## ## 1 Fig. 10 Smaller than 5 Larger than 5 38
Experiments 3 and 4: design Incongruent trial, where the cue was indicative for the prime in 4a 480ms time 67ms 33ms 33ms ##### +++++ ##### ##### 17ms ##### ##### ##### ZAAG ##### ##### ##### spin Fig. 11 Animal Object 39
Experiments 3 and 4: results 30 *** p <.001 ** p <.01 * p <.05 p <.07 ns: not sign RCE (ms) 25 20 15 10 5 0 ** ** ns * ns -5-10 ns ns ns attended unattended attended unattended Experiment 3 Experiment 4 repeated novel Fig. 12 40
Experiment 3 and 4: conclusion Significant difference between RCEs for attended VS unattended primes: strong decrease of RCEs when primes are not attended ~ Broadbent s no identification without attention account. Effects for novel word primes are weak! 41
Conclusions Study 2 Attention plays an important role in subliminal processing: when attention is divided across multiple sources of information or when the subliminal information is not attended, RCEs strongly decrease. Further support for the No identification without attention claim (Broadbent, 1958): spatial attention seems to be a necessary preliminary for subliminal semantic processing. 42
Conclusions Study 2 However: Word versus number stimuli (e.g. Damian, 2001) effects for word stimuli are less outspoken Novel versus repeated primes when using word stimuli effects for novel primes are less outspoken our Study 1 (strong effects for novel word stimuli) Explanation? - Only novel primes VS mixture 43
General conclusions Study 1: Evidence in favor of the semantic categorization hypothesis, which assumes (at least partial) semantic processing of subliminal primes Study 1: target set size and category size do not appear to be key factors in obtaining subliminal semantic priming effects Study 2: Evidence in favor of attention as a strong influencing factor in subliminal semantic priming 44
General conclusions However: The role of the nature of the primes (novel, repeated or both) remains unclear The dynamics between category size, category coherency and target set size should be further investigated The (differential?) onset and time-course of subliminal semantic priming in number VS word stimuli has never been studied Other (possibly) influencing factors (e.g. nature of task) have not yet been included These questions will be addressed in the next 2 years using new experimental research and a meta-analysis 45
Contact Info Eva Van den Bussche KULeuven - Campus Kortrijk E. Sabbelaan 53 8500 Kortrijk Belgium Tel.: +32 56 246074 Fax: +32 56 246052 Eva.Vandenbussche@Kuleuven-kortrijk.be 46