Stepping out of the Chinese Room: Word meaning with and without consciousness Roberto Bottini Daniel Casasanto Center for Mind/Brain Sciences, University of Chicago, IL University of Trento, Italy USA bottini.r@gmail.com casasanto@uchicago.edu Andrea Nadalini Davide Crepaldi International School for Advanced Studies (SISSA) Milan Center for Neuroscience Trieste, Italy Milan, Italy anadalini@sissa.it davide.crepaldi@sissa.it Italiano. Qual è il ruolo della coscienza Abstract nell’elaborazione semantica delle parole? Esperimenti di masked priming semantico English. What is the role of consciousness in mostrano che la vista di una parola può faci- language processing? Unconscious priming litare il riconoscimento di un’altra parola experiments show that words can prime other dal contenuto semantico simile (gatto – ca- words with related meanings (cat – dog), and ne). Questo effetto di priming è solitamente these priming effects are assumed to reflect interpretato come evidenza che la parola the activation of conceptual knowledge in inconscia è processata a livello semantico. semantic memory. Alternatively, however, Tuttavia, tale effetto può essere spiegato an- unconscious priming effects could reflect che sulla base di relazione tra forme lessicali predictive relationships between the words’ (senza attivazione di informazione nella me- forms, since words that are semantically re- moria semantica). Infatti, parole che sono lated are also statistically related in language semanticamente legate sono anche legate use. Therefore, unconscious “semantic” statisticamente nel linguaggio. Il priming priming effects could be due to relationships semantico inconscio potrebbe semplicemente between words’ forms mimicking conceptual emulare relazioni concettuali, come nel fa- relationships, as in Searle’s Chinese Room moso esperimento mentale della stanza cine- thought experiment. To distinguish word- se di Searle. Per distinguere il priming lessi- form-based and semantics-based accounts of cale dal priming semantico abbiamo condot- priming we conducted an experiment in to un esperimento in cui parole temporali which temporal words (e.g., earlier, later) (ieri, domani) erano precedute da parole were preceded by spatial words that were spaziali mostrate sia a livello subliminale processed either consciously or unconscious- che supraliminale. Il tempo è tipicamente ly. Time is typically conceptualized as a spa- concettualizzato attraverso mappe spaziali tial continuum extending along either the che si estendono lungo l’asse sagittale (il sagittal (front-back) or the lateral (left-right) passato è dietro, e il futuro davanti) e lungo axis, but only the sagittal space-time map- l’asse laterale (il passato è a sinistra, futuro ping is encoded in language (e.g. the future is a destra). Solo la mappatura sagittale è però ahead, not to the right). Results showed that codificata nel linguaggio (il futuro è davanti, temporal words were primed both by sagittal non a destra). I risultati mostrano come sia words (back, front) and lateral words (left, le parole sagittali (dietro, davanti) che quelle right) when primes were perceived con- laterali (sinistra, destra) facilitano l'elabora- sciously, as predicted by both wordform- zione di parole temporali (ad esempio prima based and semantics-based accounts. Yet, e dopo), quando percepite consciamente. Al only sagittal words produced an unconscious contrario, quando i prime sono elaborati al priming effect, as predicted by the word- di fuori della coscienza, l’effetto sull’asse form-based account. Unconscious word pro- laterale viene meno. Il processo inconscio cessing appears to be limited to relationships delle parole sembra dunque essere limitato a between words’ forms, and consciousness relazioni tra forme lessicali; la coscienza may be needed to activate words’ meanings. potrebbe essere necessaria per attivarne il may be alternatively explained through predic- significato. tive relationships between words’ forms estab- lished in language use. According to the word- 1 Introduction form-based account, cat would prime dog simply because the two words share a similar contextual What role does consciousness play in word distribution. No conceptual representation is in- meaning’s construction? As previous literature volved, as the locus of the unconscious semantic has pointed out, lexical items seem to be pro- priming would be the lexical system, not seman- cessed up to the semantic level even when pro- tic memory (Collins & Loftus, 1975). cessed out of awareness (Quinn & Kinoshita, This latter interpretation of unconscious word 2008; Ansorge, Kiefer, Khalid, Grassl, & König, processing somehow resembles the Chinese 2010). Most evidence for this claim comes from Room thought experiment developed by the phi- masked priming: When two words are sequen- losopher John Searle, where an English-speaking tially presented, the recognition of the latter is man is closed in a room receiving message writ- made easier if the two are semantically related ten in Chinese characters (Searle, 1980). Due to a (cat-dog), even when the visibility of the former set of norms that determine the relationships be- (the prime) is prevented by displaying it very tween those characters (if you see X followed by briefly, embedded between visual masks Y, than reply Z), he is able to provide answers (Forster, 2006; Dehaene et al., 1998). For in- that would look perfectly appropriate to a native stance, participants are likely to classify more speaker. From the outside, it would appear that quickly the word dog as referring to a living enti- the man has a good understanding of the lan- ty when it is preceded by the semantically related guage, while instead he is acting on the basis of word cat, rather than by a semantically unrelated associations between word forms. Indeed, if he word like apple. As similar effects are attested received a message saying that the room is about when the prime word is clearly visible, it has to explode, he would reply appropriately; but been suggested that lexical items can be pro- would not leave the room. cessed up to the semantic level irrespective of their visibility. We will refer to this perspective For the semantic-based and wordform-based as the semantic-based account of masked prim- accounts of masked priming to be distinguished, ing, as it assumes that words are processed be- it is necessary to find concepts that are related in yond their surface structure and activate concep- the semantic system, but not in the lexical sys- tual knowledge about their referents. Such tem. This is the case for the metaphorical rela- knowledge is thought to be stored within the se- tionship linking time to space. The two domains mantic memory, where concepts and concepts’ are strictly intertwined in the human mind, in features are represented in an interconnected such a way that space is often used to think about network (Tulving, 1972; Masson, 1995). time (Lakoff & Johnson, 1980). Time conceptu- In such a view, cat would prime dog as both alization involves both the sagittal and the lateral words refer to mammals that have four legs, have axis (Casasanto & Bottini, 2014; Bonato, Zorzi a tail, can be pet, and so on; and therefore they & Umiltà, 2012). For example, participants are are more closely related to each other than to faster in responding to past-related words by apple. providing a leftward response, and to future- However, there is an aspect of semantic similari- related words by providing a rightward response, ty that has been largely overlooked in the prim- relative to the opposite pattern. The same holds ing-related literature, that words with similar for the sagittal arrangement, with backward re- meaning tend to have a similar contextual distri- sponse associated with past-related words and bution. As corpus-based studies have pointed forward response associated with future-related out, words referring to entities with similar per- words. Moreover, neurological evidence shows ceptual and conceptual attributes tend be used that patients with hemispatial neglect have also together (e.g., dog and cat are more likely to co- impairments in temporal judgments (i.e. if they occur in the same sentence than dog and apple), neglect the left side on space, they also show and to be used in similar contexts (e.g., both cat worst memory for past-related events; Saj, and dog tend to appear when speaking about Fuhrman, Vuilleumier, & Boroditsky, 2013). pets, whether or not they co-occur within a given Finally, people have been found to use hand ges- utterance; Louwerse, 2011; Landauer & Dumais, tures along both the lateral and the sagittal line 1997). Based on this fact, unconscious priming when speaking about time (Casasanto & Jasmin, 2012). Critically, while the sagittal mapping is Results and discussion: analyses were conducted linguistically encoded in sentences such as “a only on “GO” trials. Inaccurate trials (less than bright future in front of you”, the lateral one is 1%) were excluded. In order to reduce the effect not. The existence of these two mental timelines, of outliers, those individual datapoints standing and the fact that only one of them is linguistical- at more than 2 standard deviations from each ly expressed, offers the ideal test-bed for con- participant’s mean (~5% of the correct trials) trasting the semantic-based and wordform-based were also removed from the analyses. A 2–by–2 accounts of unconscious word processing. The ANOVA on the log-transformed RTs revealed a latter predicts that space–time priming would significant main effect of Congruity, F(1, 59)= only emerge along the linguistically encoded 11.47, p= 0.001, indicating that participants were sagittal axis when primes are kept outside of faster in congruent trials (535 ms) compared to awareness; while the former would predict prim- incongruent ones (540 ms). We found no effect ing to emerge along both axes, both supra- and of Axis, F(1, 59)= 0.41, p> .250, and no Axis by subliminally. Congruity interaction, F(1, 59)= 0.06, p>0.250. We tested these predictions in a priming study Pairwise comparisons showed that the priming with spatial words related to the lateral (left- effect was significant both in the sagittal (4 ms; right) and the sagittal (ahead-behind) axis as F(1, 59)= 5.79, p= 0.02) and the lateral axis (6 primes, and temporal words referring to either ms; F(1, 59)= 6.76, p= 0.01). the past (yesterday) or the future (tomorrow) as Thus, significant congruity effects were pro- target stimuli. In the first experiment, primes duced both by sagittal and lateral spatial prime were clearly visible In the second experiment, words, consistent with previous studies that pro- prime visibility was prevented by means of a vide evidence for sagittal and lateral mental masking procedure. timelines. 2 Experiments 2.2 Experiment 1b - Subliminal primes 2.1 Experiment 1a - Visible primes Participants: 60 volunteers from the same popu- lation as in Experiment 1a were recruited into the Participants: 60 volunteers were recruited for the experiment. None of them took part in the previ- experiment; all subjects were right-handed, and ous experiment. they all stated being native Italian speakers, with Stimuli, apparatus and procedure were the same normal or corrected-to-normal vision and no his- as in Experiment 1a with one exception, i.e the tory of neurological disorders. Each subject gave blank screens that were displayed before and af- written informed consent for participation. ter the prime word were replaced with two visual Stimuli, apparatus and procedure: Primes were 2 masks in order to make the prime invisible (sub- spatial words related to the lateral axis (“sinis- jects were not informed of the presence of the tra”, left, and “destra”, right) and 2 spatial words primes). related to the sagittal axis (“davanti”, front, and Prime visibility task: after the end of the last part “dietro”, back). Target stimuli were 8 temporal of the experiment, all subjects were informed words. Four of them refer to the past (“prima”, about the presence of the prime word between earlier, “ieri”, yesterday, “passato”, past, the masks. Then, they performed a prime visibil- “scorso”, previous), and four refers to the future ity test that included 10 practice and 128 experi- (“dopo”, later, “domani”, tomorrow, “futuro”, mental trials. The stimuli to be detected were the future, “successivo”, next). same spatial words we used in the previous ex- Each trial consisted of a fixation point (+) dis- periment in half of the trials, and a string of iden- played for 750 ms. Then a blank screen was tical lowercase letters () in the other shown for 200ms, followed by the prime and by half. another blank screen, both lasting 50 ms. Finally, Results and discussion: only the “GO” trials, in the target word was presented for 1500 ms, or which participants provided a response, were until a response was provided. analyzed. Inaccurate trials (less than 1%) were Participants engaged in a GO/NO_GO task: They excluded. In order to reduce the effect of ex- had to press a key if the target word referred to tremely long and short RTs, those individual the past and do nothing if the target word re- datapoints standing at more than 2 standard devi- ferred to the future, or vice versa, according to ations from each participant’s mean (~4% of the the block instructions. correct trials) were also removed from the anal- yses. A 2–by–2 ANOVA on the log-transformed RTs cognition with subliminal words. Cognition, revealed a significant main effect of Congruity, 116(3),303–320. F(1, 58)= 27.63, p< .001, and an Axis by Con- Bonato, M., Zorzi, M., & Umiltà, C. (2012). When gruity interaction, F(1, 58)= 14.986, p< 0.001, time is space: evidence for a mental time line. which we followed up through pairwise compari- Neuroscience and Biobehavioral Reviews, 36(10), sons showing that priming was significant for the 2257–73. sagittal axis (9 ms; F(1, 58)= 40.21, p< 0.001), Casasanto, D., & Bottini, R. (2014). Mirror Reading but not for the lateral axis (2 ms; F(1, 58)= 1.52, Can Reverse the Flow of Time. Journal of p= 0.22). Experimental Psychology. General, 143(2). No participant reported having noticed the prime. Casasanto, D., & Jasmin, K. (2012). The Hands of Overall, the average d-prime value was 0.33 Time: Temporal gestures in English speakers. (SD= 0.37). Although significantly different Cognitive Linguistics, 23(4), 643 – 674. from zero, t(58)= 7.03, p< 0.001, this value is widely taken to indicate that the prime was effec- Collins, A. M., & Loftus, E. F. (1975). A Spreading- tively masked from perceivers’ awareness Activation Theory of Semantic Processing, Psych Review, 82(6), 407–428. (Kouider & Dupoux, 2005). Experiment 1b clearly suggests that spatio– Dehaene, S., Naccache, L., Le Clec’H, G., Koechlin, temporal masked priming is limited to the sagit- E., Mueller, M., Dehaene-Lambertz, G., Van de tal axis, with no apparent effect on the lateral Mortele, P-F, & Le Bihan, D. (1998). Imaging axis. Thus, the pattern of results provides evi- unconscious semantic priming. Nature, 395(6702), 597–600. dence in favor of the wordform-based account of unconscious word processing. Fodor, J. (1983). The modularity of mind. Cambridge. MIT Press. 3 Conclusion Forster, K. I. (2006). Early activation of category in- In this study we looked at the nature of word formation in visual word recognition More on the turple effect, The Mental Lexicon, 1, 35–58. processing with and without awareness. Using the relationships between space and time, we Kouider, S., & Dupoux, E. (2005). Subliminal speech were able to disentangle the wordform-based priming. Psychological Science, 16(8), 617–25. from the semantic-based account of masked Lakoff, G., & Johnson, M. (1980). Metaphors we live priming. When words were clearly visible, we by. Chicago, IL: University of Chicago Press. found priming effect on both axes, which reflects Landauer, T. K., & Dumais, S. T. (1997). A solution the sagittal, linguistically encoded, timeline, as to Plato ’ s problem : The Latent Semantic Analy- well as the lateral mapping, which relies only on sis Theory of Acquisition , Induction , and Repre- conceptual knowledge. Conversely, subliminal sentation of Knowledge. Psychological Review, priming was obtained only with the sagittal 104(2), 211–240. words, matching the predictions of the word- Louwerse, M. M. (2011). Symbol interdependency in form-based account. Therefore, our data suggests symbolic and embodied cognition. Topics in Cog- that when people read words unconsciously, ac- nitive Science, 3(2), 273–302. tivation spreads only between predictively relat- ed wordforms. Masson, M. E. J. (1995). A Distributed Memory Unconscious priming between semantically re- Model for Semantic Priming. Journal of Experi- mental Psychology. lated words may mimic semantic priming, much in the same way as the man inside Searle’s Chi- Quinn, W. M., & Kinoshita, S. (2008). Congruence nese Room mimics knowledge of Chinese, on the effect in semantic categorization with masked basis of “meaningless” wordform-wordform rela- primes with narrow and broad categories. Journal tionships. Unconscious word processing appears of Memory and Language, 58(2), 286–306. to be limited to relationships between words’ Saj, A., Fuhrman, O., Vuilleumier, P., & Boroditsky, forms, and consciousness may be needed to acti- L. (2013). Patients With Left Spatial Neglect Also vate words’ meanings. Neglect the “Left Side” of Time. Psychological Science, (November). Searle, J. R. (1980). Minds , brains , and programs, Reference BBS, 3, 1–19. Ansorge, U., Kiefer, M., Khalid, S., Grassl, S., & Kö- nig, P. (2010). Testing the theory of embodied Tulving, E. (1972). Episodic and semantic memory. In Organization of memory (Vol. 381). London: Aca- demic.