Learning Recursion Bob Frank Department of Linguistics Yale University Saturday, January 7, 2012 1 Recursion and human language Hauser, Chomsky and Fitch (2002) “We hypothesize that FLN only includes recursion and is the only uniquely human component of the faculty of language.” Saturday, January 7, 2012 2 Recursion and Human Language • What counts as recursion? S S a b S a S b S abS S aSb → a b S → a S b S ab S ab → a b → a b Saturday, January 7, 2012 3 R EPORTS Finally, the observation that the SRP 21. A. Perrakis, T. K. Sixma, K. S. Wilson, V. S. Lamzin, Acta and S. R. Sprang for critical reading of this manu- GTPases behave as reciprocal GTPase acti- Crystallogr. D53, 448 (1997). script, and Panchika Prangkio for her contribution. 22. Sequence motifs are defined for T. aquaticus Ffh and This work was supported by grant GM58500 from the vating proteins (5) can now be understood to FtsY in table S1. NIH. Portions of this work were performed at the be a consequence of the formation of a shared 23. J. A. Newitt, H. D. Bernstein, Eur. J. Biochem. 245, 720 DuPont-Northwestern-Dow Collaborative Access catalytic chamber between them. Indeed, the (1997). Team (DND-CAT ) Synchrotron Research Center, Sec- 24. D. M. Freymann, R. J. Keenan, R. M. Stroud, P. Walter, tor 5, and BioCARS, Sector 14, of the Advanced reciprocal hydrogen bonding between the Nat. Struct. Biol. 6, 793 (1999). Photon Source (APS) at Argonne National Laboratory. bound nucleotides may itself be catalytically 25. U. D. Ramirez et al., J. Mol. Biol. 320, 783 (2002). Use of the APS was supported by the U.S. Depart- important. However, the structure of the com- 26. K. Scheffzek et al., Science 277, 333 (1997). ment of Energy, Basic Energy Sciences, Office of plex also demonstrates how the initial engage- 27. J. J. Tesmer, D. M. Berman, A. G. Gilman, S. R. Sprang, Science, under Contract No. W-31-109-Eng-38. Cell 89, 251 (1997). DND-CAT was supported by DuPont, Dow, the state ment of the two proteins can function as a 28. P. J. Rapiejko, R. Gilmore, Mol. Biol. Cell 5, 887 of Illinois, and the NSF. Use of the BioCARS was latch, in that a number of structural elements, (1994). supported by NIH, National Center for Research Re- 29. T. Connolly, P. J. Rapiejko, R. Gillmore, Science 252, sources, under grant number RR07707. Support from including the bound nucleotides, contribute to the R. H. Lurie Comprehensive Cancer Center of an intricate interface that is unlikely to disso- 1171 (1991). 30. E. de Leeuw et al., EMBO J. 19, 531 (2000). Northwestern University to the Structural Biology Uniquely human?Facility is acknowledged. Coordinates and structure ciate until two subsequent steps, signal pep- 31. J. S. Millman, H. Y. Qi, F. Vulcu, H. D. Bernstein, D. W. tide transfer followed by nucleotide hydroly- Andrews, J. Biol. Chem. 276, 25982 (2001). factors have been deposited with the Protein Data Bank and assigned accession code 1OKK. sis, occur. This kind of mechanism for the SRP 32. J. D. Miller, H. D. Bernstein, P. Walter, Nature 367, 657 (1994). Supporting Online Materials GTPases is consistent with a process requiring 33. S. R. Sprang, Annu. Rev. Biochem. 66, 639 (1997). www.sciencemag.org/cgi/content/full/303/5656/373/ assembly of multiple components, and it can be 34. H. Lu¨tcke, S. High, K. Ro¨misch, A. J. Ashford, B. DC1 distinguished from one in which the GTPases Dobberstein, EMBO J. 11, 1543 (1992). Materials and Methods Figs. S1 to S6 act along a signaling pathway. Extending the 35. R. M. Cleverley, L. M. Gierasch, J. Biol. Chem. 277, 46763 (2002). Tables S1 and S2 metaphor, the GTP molecules themselves can 36. P. F. Egea et al., Nature 427, 215 (2004). References be imagined as “explosive bolts” in that they are 37. We thank C. W. Carter, S. McGovern, U. D. Ramirez, 26 August 2003; accepted 20 November 2003 integral to the interface that holds the proteins together, and so promote transfer of the translat- ing ribosomal cargo, but that they also provide, Computational Constraints on by their hydrolysis, the “explosion” that dis- engages the components of the targeting com- plex (fig. S6). This conception of the role of Syntactic Processing in a on February 27, 2009 GTP is somewhat distinct from the classic GTPase switch model and provides insight into Nonhuman Primate the logic of the SRP GTPases that may be rel- 1 2 evant to understanding other GTPases that func- W. Tecumseh Fitch * and Marc D. Hauser tion in the assembly of cellular components. Note added in proof: A structure of a The capacity to generate a limitless range of meaningful expressions from a similar complex of the SRP GTPases in a finite set of elements differentiates human language from other animal com- different crystal form was independently de- munication systems. Rule systems capable of generating an infinite set of termined and is reported by Egea et al.(36). outputs (“grammars”) vary in generative power. The weakest possess only local www.sciencemag.org organizational principles, with regularities limited to neighboring units. We used References and Notes a familiarization/discrimination paradigm to demonstrate that monkeys can 1. R. J. Keenan, D. M. Freymann, R. M. Stroud, P. Walter, spontaneously master such grammars. However, human language entails more Annu. Rev. Biochem. 70, 755 (2001). sophisticated grammars, incorporating hierarchical structure. Monkeys tested 2. H. D. Bernstein et al., Nature 340, 482 (1989). 3. K. Ro¨misch et al., Nature 340, 478 (1989). with the same methods, syllables, and sequence lengths were unable to master 4. J. D. Miller, H. Wilhelm, L. Gierasch, R. Gilmore, P. a grammar at this higher, “phrase structure grammar” level. Walter, Nature 366, 351 (1993). 5. T. Powers, P. Walter, Science 269, 1422 (1995). Downloaded from 6. Q. A. Valent et al., EMBO J. 17, 2504 (1998). Syntax is one key component of human lan- far beyond the simple concatenation proce- 7. C. Moser, O. Mol, R. S. Goody, I. Sinning, Proc. Natl. guage, with no known equivalent in animal dures sometimes called “syntax” in animal Acad. Sci. U.S.A. 94, 11339 (1997). communication systems. The limitless ex- communication (1–3). However, the evolu- 8. P. Peluso, S. O. Shan, S. Nock, D. Herschlag, P. Walter, Biochemistry 40, 15224 (2001). pressive power of human language requires tion of the language faculty presumably in- 9. Y. Lu et al., EMBO J. 20, 6724 (2001). structures, termed phrases or sentences, volved the incorporation of some ancestral 10. S. Padmanabhan, D. M. Freymann, Structure 9, 859 above the word level (or, by analogy, above primate cognitive capabilities. Thus, a critical (2001). 11. P. J. Rapiejko, R. Gilmore, Cell 89, 703 (1997). the single call level in animals). Linguistic question is whether hierarchical processing 12. W. Song, D. Raden, E. C. Mandon, R. Gilmore, Cell syntax involves the rearrangement and per- was one of these preexisting abilities, perhaps 100, 333 (2000). mutation of such abstract hierarchical struc- evolved to serve noncommunicative func- 13. D. M. Freymann, R. J. Keenan, R. M. Stroud, P. Walter, Nature 385, 361 (1997). tures,Saturday, often January with 7, concomitant 2012 changes in tions (e.g., motor control, number, or social 4 14. P. J. Focia, H. Alam, T. Lu, U. D. Ramirez, D. M. meaning. The production and perception of cognition) (4–12). Freymann, Proteins 54, 222 (2004). these hierarchical syntactic structures is a Rule systems capable of generating infi- 15. G. Montoya, C. Svensson, J. Luirink, I. Sinning, Nature 385, 365 (1997). core capability underlying human linguistic nite sets of sequences (“grammars”) are ar- 16. G. Montoya, K. Kaat, R. Moll, G. Scha¨fer, I. Sinning, competence. This level of organization goes ranged in a mathematical hierarchy of in- Structure 8, 515 (2000). creasing generative power, termed the 17. I. R. Vetter, A. Wittinghofer, Science 294, 1299 Chomsky hierarchy (13, 14). The weakest (2001). 1School of Psychology, University of St. Andrews, St. 18. I. V. Shepotinovskaya, D. M. Freymann, Biochim. Bio- Andrews, Fife, KY16 9AJ, Scotland. 2Department of class in this hierarchy are finite state gram- phys. Acta 1597, 107 (2002). Psychology, Harvard University, Cambridge, MA mars (FSGs), which can be fully specified by 19. I. V. Shepotinovskaya, P. J. Focia, D. M. Freymann, 02138, USA. transition probabilities between a finite num- Acta Crystallogr. D59, 1834 (2003). 20. Materials and methods are available as supporting *To whom correspondence should be addressed. ber of “states” (e.g., corresponding to words material on Science Online. E-mail: [email protected] or calls). Recent evidence suggests that pars- www.sciencemag.org SCIENCE VOL 303 16 JANUARY 2004 377 R EPORTS ing procedures at this superficial level of lematic (20, 25). Because limited output from sen at random. Crucially, syllables for each complexity are spontaneously available to a PSG can always be approximated by a more class were sampled without replacement, both human infants and nonhuman primates complicated FSG (at the limit, a memorized because otherwise the possibility of exact (3, 15–19). However, FSGs are inadequate to list of exemplars), it is difficult to prove acoustic repetitions in the PSG and not in generate all the structures of any human lan- conclusively that subjects have learned the the FSG would make the two grammars guage (13, 20), because all languages mini- former. This is equally true for human or distinguishable on superficial grounds. The R EPORTS mally require procedures at the next level of animal subjects. However, failure to master a FSG was (AB)n,inwhicharandom“A” ing procedures at this superficial level ofcomplexity,lematic termed (20, 25).