Professional Documents
Culture Documents
Language and Cognition: Frontiers in Behavioral Neuroscience December 2014
Language and Cognition: Frontiers in Behavioral Neuroscience December 2014
Language and Cognition: Frontiers in Behavioral Neuroscience December 2014
net/publication/286313223
CITATIONS READS
14 1,531
2 authors, including:
Leonid Perlovsky
Harvard University
333 PUBLICATIONS 6,359 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Leonid Perlovsky on 16 April 2016.
FUNCTION OF LANGUAGE AND 1998). This is known as combinatorial of situations and abstract concepts ini-
COGNITION IN THINKING complexity, CC. This difficulty in mod- tially exist in vague states. Throughout
Do we think with language, or is it just eling the mind has been overcome by the rest of life, language guides acqui-
a communication device used for expres- dynamic logic (Perlovsky, 2001, 2006a,b, sition of cognitive representations from
sion of completed thoughts? What is a 2007a; Perlovsky et al., 2011). Whereas experience. Vague cognitive representa-
difference between language and cogni- classical logic considers static statements tions become more crisp and concrete.
tion? Chomsky (1995) suggested that these such as “this is a chair,” dynamic logic Thinking involves both language and cog-
two abilities are separate and indepen- models processes from vague to crisp rep- nition, and as we discuss later thinking
dent. Cognitive linguistics emphasizes a resentations. These processes do not need about abstract ideas usually involves lan-
single mechanism for both (Croft and to consider combinations, an initial vague guage more than cognition, not too differ-
Cruse, 2004). Evolutionary linguistics con- state of a “chair” matches any object in the ent from thinking by children.
siders the process of transferring language field of view, and at the end of the pro-
from one generation to the next one cess it matches the chair actually present, THE DUAL HIERARCHY
(Cangelosi and Parisi, 2002; Christiansen without CC. Cognitive representations are organized
and Kirby, 2003; Hurford, 2008). This The second difficulty is similar still in mind in an approximate hierarchy
process is a “bottleneck” that forms the even more complex. It is related to the (Grossberg, 1988) from sensor-motor per-
language. Brighton et al. (2005) demon- fact that “events” and “situations” in the cepts near “bottom,” to objects “higher
strated emergence of compositional lan- world do not necessarily exist “ready for up,” to situations, and to still more abstract
guage due to this bottleneck. Still, none cognition.” There are many combinations cognitive representations. Language rep-
of these approaches resulted in a com- of percepts and objects, a near infin- resentations are organized in a parallel
putational theory explaining how humans ity, events and situations important for hierarchy from sounds, and words for
acquire language and cognition. Here I understanding and learning have to be objects and situations, to phrases, and to
discuss a computational model overcom- separated from those that are just ran- more abstract language representations.
ing previous difficulties and based on a dom collections of meaningless percepts Our previous discussion can be described
hypothesis that language and cognition are or random objects (Perlovsky and Ilin, by an integrated mathematical model of
two separate and closely integrated abili- 2012). Events and situations recognized language and cognition forming a dual
ties. I identify their functions and discuss by non-human animals are very limited hierarchy (Perlovsky, 2009a), as illustrated
why human thinking ability requires both compared to human abilities to differenti- in Figure 1. Neural evidence suggests that
language and cognition. ate events in the world. Human cognitive the hierarchy is approximate, not as defi-
Among fundamental mechanisms of abilities acquire their power due to lan- nite as shown in this figure.
cognition are mental representations, guage. Language is “easier” to learn than Hierarchical organization of cogni-
memories of objects and events (Perlovsky, cognitive representations. Language rep- tion and related brain structures are
2001, 2006a). The surrounding world is resentations: words, phrases exist in the reviewed in (Badre, 2008). In particular,
understood by matching mental repre- surrounding language “ready made,” cre- anterior-posterior axis corresponds to a
sentations to patterns in sensor signals. ated during millennia of cultural evolu- gradient of abstract-concrete cortex func-
However, mathematical modeling of this tion. Therefore, language could be learned tions. Hierarchical organization of lan-
process since the 1950s met with difficul- without much real-life experience; only guage functions is also well established.
ties. The first difficulty is related to a need interactions with language speakers are However, hierarchical organization of lan-
to consider combinations of sensor signals, required. Every child learns language early guage does not correspond to a par-
objects, and events. The number of com- in life before acquiring full cognitive ticular spatial axis in the brain, it is
binations is very large and even a limited understanding of events and their cogni- distributed (Price, 2012). Therefore, the
number of signals or objects form a very tive meanings. Thus, language is learned dual hierarchy in Figure 1 is a func-
large number of combinations, exceeding early in life with only limited cognitive tional hierarchy not organized along a
all interactions of all elementary particles understanding of the world (Perlovsky, spatial axis in the brain as in this fig-
in a lifetime of the Universe (Perlovsky, 2009a, 2012c). Cognitive representations ure. A fundamental aspect of acquiring
of emotional connections between sound Binder, J. R., Westbury, C. F., McKiernan, K. (Heidelberg: Springer Verlag), 73–108. doi:
and meaning depends on language inflec- A., Possing, E. T., and Medler, D. A. (2005). 10.1007/978-3-540-73267-9_5
Distinct brain systems for processing concrete and Perlovsky, L. I. (2007b). Evolution of languages, con-
tions. In particular, after English lost most
abstract concepts. J. Cogn. Neurosci. 17, 1–13. doi: sciousness, and cultures. IEEE Comput. Intell. Mag.
of its inflections, it became a low emo- 10.1162/0898929054021102 2, 25–39. doi: 10.1109/MCI.2007.385364
tional language, powerful for science and Brighton, H., Smith, K., and Kirby, S. (2005). Perlovsky, L. I. (2009a). Language and cog-
engineering. At the same time English is Language as an evolutionary system. Phys. Life Rev. nition. Neural Netw. 22, 247–257. doi:
losing autonomous connections to cultural 2, 177–226. doi: 10.1016/j.plrev.2005.06.001 10.1016/j.neunet.2009.03.007
Cangelosi, A., and Parisi, D. (eds.). (2002). Simulating Perlovsky, L. I. (2009b). Language and emotions:
values that used to be partially inherent in emotional sapir-whorf hypothesis. Neural Netw.
the Evolution of Language. London: Springer. doi:
language sounds. Fast change of cultural val- 10.1007/978-1-4471-0663-0 22, 518–526. doi: 10.1016/j.neunet.2009.06.034
ues during recent past is usually attributed Chomsky, N. (1995). The Minimalist Program. Perlovsky, L. I. (2010). Musical emotions: functions,
to progress in thinking, whereas effects of Cambridge: MIT Press. origin, evolution. Phys. Life Rev. 7, 2–27. doi:
change in emotionality of language sounds Christiansen, M. H., and Kirby, S. (2003). Language 10.1016/j.plrev.2009.11.001
Evolution. New York, NY: Oxford University Press. Perlovsky, L. I. (2012a). Cognitive function, origin,
have not been noticed. doi: 10.1093/acprof:oso/9780199244843.001.0001 and evolution of musical emotions. Music. Sci. 16,
Emotional prosody can be impor- Croft, W., and Cruse, D. A. (2004). Cognitive 185–199; doi: 10.1177/1029864912448327
tant for overcoming cognitive dissonance. Linguistics. Cambridge: Cambridge University Perlovsky, L. I. (2012b). Emotionality of languages
Cognitive dissonance is a discomfort Press. doi: 10.1017/CBO9780511803864 affects evolution of cultures. Rev. Psychol. Front. 1,
Deacon, T. W. (1997). The Symbolic Species: The Co- 1–13.
due to holding contradictory cognitions
Evolution of Language and the Brain. New York, NY: Perlovsky, L. I. (2012c). Brain: conscious and uncon-
(Festinger, 1957; Harmon-Jones et al., Norton. scious mechanisms of cognition, emotions, and
2009). It is resolved by discarding con- Festinger, L. (1957). A Theory of Cognitive Dissonance. language. Brain Sci. 2, 790–834.
tradictions. If a new word contradicts Stanford, CA: Stanford University Press. Perlovsky, L. I. (2012d). Cognitive function
existing knowledge its meaning might Grossberg, S. (1988). Neural Networks and of music part, I. Interdisc. Sci. Rev. 7,
Natural Intelligence. Cambridge: MIT 129–142.
be discarded. Emotional prosody as well
Press. Perlovsky, L. I. (2013a). A challenge to human evolu-
as songs could be fundamental mecha- Harmon-Jones, E., Amodio, D. M., and Harmon- tion – cognitive dissonance. Front. Psychol. 4:179.
nisms that overcome cognitive dissonance Jones, C. (2009). “Action-based model of disso- doi: 10.3389/fpsyg.2013.00179
and enable keeping new contradictory nance: a review, integration, and expansion of Perlovsky, L. I. (2013b). Cognitive function of music
knowledge (Masataka and Perlovsky, 2012; conceptions of cognitive conflict,” in Adv Exp part II. Interdisc. Sci. Rev. 38, 149–173.
Soc Psychol 41, ed M. P. Zanna (Burlington, ON: Perlovsky, L. I., Deming, R. W., and Ilin, R.
Perlovsky, 2013a). (2011). Emotional Cognitive Neural Algorithms
Academic Press), 119–166.
Hauser, M. D., Chomsky, N., and Fitch, W. T. (2002). with Engineering Applications. Dynamic Logic:
CONCLUSION AND EXPERIMENTAL The faculty of language: what is it, who has it, and from vague to crisp. Heidelberg: Springer. doi:
PREDICTIONS how did it evolve? Science 298, 1569–1579. doi: 10.1007/978-3-642-22830-8
10.1126/science.298.5598.1569 Perlovsky, L. I., and Ilin, R. (2010). Neurally and
This article advances a hypothesis about
Hurford, J. (2008). “The evolution of human com- mathematically motivated architecture for lan-
functions of language and cognition in munication and language,” in Sociobiology of guage and thought. Open Neuroimag. J. 4, 70–80.
thinking, and possible model of their Communication: an Interdisciplinary Perspective, Perlovsky, L. I., and Ilin, R. (2012). Mathematical
interactions. This is the only computable eds P. D’Ettorre and D. Hughes (New York, NY: model of grounded symbols: perceptual sym-
model explaining a number of myster- Oxford University Press), 249–264. bol system. J. Behav. Brain Sci. 2, 195–220. doi:
Levine, D. S., and Perlovsky, L. I. (2008). 10.4236/jbbs.2012.22024
ies about language and cognition and Price, C. J. (2012). A review and synthesis of the first
Neuroscientific insights on biblical myths:
overcoming computational difficulties. It simplifying heuristics versus careful think- 20 years of PET and fMRI studies of heard speech,
makes a number of predictions that ing: scientific analysis of millennial spiritual spoken language and reading. Neuroimage 62,
could be experimentally tested, including issues. Zygon J. Sci. Relig. 43, 797–821. doi: 816–847. doi: 10.1016/j.neuroimage.2012.04.062
the following: cognitive representations 10.1111/j.1467-9744.2008.00961.x Tversky, A., and Kahneman, D. (1974). Judgment
Masataka, N., and Perlovsky, L. I. (2012). The efficacy under uncertainty: heuristics and biases.
model the world, while language repre- Science 185, 1124–1131. doi: 10.1126/sci-
of musical emotions provoked by Mozart’s music
sentations only model language; abstract for the reconciliation of cognitive dissonance. Sci. ence.185.4157.1124
cognitive representations can only be Rep. 2, 694. doi: 10.1038/srep00694
acquired due to language; abstract cogni- Perlovsky, L. I. (1998). Conundrum of combinatorial Received: 26 August 2013; accepted: 02 September 2013;
complexity. IEEE Trans. PAMI 20, 666–670. doi: published online: 19 September 2013.
tion is more clearly represented in lan-
10.1109/34.683784 Citation: Perlovsky L (2013) Language and cognition—
guage whereas cognitive representations Perlovsky, L. I. (2001). Neural Networks and Intellect: joint acquisition, dual hierarchy, and emotional
may remain vague throughout life. Using Model-Based Concepts. New York, NY: prosody. Front. Behav. Neurosci. 7:123. doi: 10.3389/
Oxford University Press. fnbeh.2013.00123
ACKNOWLEDGMENTS Perlovsky, L. I. (2006a). Toward physics of the mind: This article was submitted to the journal Frontiers in
concepts, emotions, consciousness, and symbols. Behavioral Neuroscience.
I am thankful for discussions with my Copyright © 2013 Perlovsky. This is an open-access
Phys. Life Rev. 3, 22–55. doi: 10.1016/j.plrev.2005.
colleagues, Michel Cabanac and Nobuo 11.003 article distributed under the terms of the Creative
Masataka. Perlovsky, L. I. (2006b). Fuzzy dynamic logic. New Commons Attribution License (CC BY). The use, dis-
Math. Nat. Comput. 2, 43–55. doi: 10.1142/ tribution or reproduction in other forums is permitted,
REFERENCES S1793005706000300 provided the original author(s) or licensor are credited
Badre, D. (2008). Cognitive control, hierarchy, Perlovsky, L. I. (2007a). “Neural dynamic logic and that the original publication in this journal is cited,
and the rostro–caudal organization of the of consciousness: the knowledge instinct,” in in accordance with accepted academic practice. No use,
frontal lobes. Trends Cogn. Sci. 12, 193–200. doi: Neurodynamics of Higher-Level Cognition and distribution or reproduction is permitted which does not
10.1016/j.tics.2008.02.004 Consciousness, eds L. I. Perlovsky and R. Kozma comply with these terms.