Sindbad~EG File Manager

Current Path : /usr/home/beeson/public_html/michaelbeeson/research/papers/
Upload File :
Current File : /usr/home/beeson/public_html/michaelbeeson/research/papers/ccc.dvi

����;� TeX output 2007.01.24:1951������Q`@�������'`@���%��
src:35ccc.tex�D��tG�G�cmr17�Constructivit��qy���V,��TComputabilit�y�,��Tand�the�Con��qtin�uum���������7��X�Qcmr12�Mic��rhael��Beeson������+�N�ffcmbx12�Abstract�����
src:38ccc.tex�K�`y

cmr10�The��nature�of�the�con���tin�uum��has�long�b�Geen�an�imp�ortan���t�issue�in�the�founda-���tions��dof�mathematics.�It�pla���y�ed��dan�imp�Gortan���t�role�in�the�w�ork�of�Dedekind,�W��*�eyl,���and���Brou���w�er,�as�w�ell�as�early�axiomatic�geometers.�When�recursiv�e�function�the-���ory�$�w���as�dev�elop�Ged,�it�w�as�immediately�applied�to�the�con�tin�uum,�via�T��*�uring's���\computable"��;n���um�b�Gers.�But�then�the�construction�of�\singular�co�v�ers"�sho�w�ed,���it�I�seemed,�that�the�recursiv���e�reals�had�measure�zero!�Do�Ges�this�mean�that�w�e���ha���v�e�:\to�c���ho�Gose�b�et���w�een�:\constructiv�e�reasoning�consisten�t�with�Ch�urc�h's�thesis���and���a�unit�in���terv��q�al�of�p�Gositiv�e�measure?�Bishop's�measure�theory�sa�v�ed�us�from���the��
horns�of�this�dilemma{or�did�it�just�sw���eep�the�dicult�y�under�the�carp�Get?�W��*�e���study�
�the�relationship�b�Get���w�een�
�Bishop's�measure�theory�and�the�recursiv���e�singu-���lar�6Gco���v�ers;�as�a�result�of�this�analysis,�w�e�iden�tify�a�logical�principle�FP�6?(fullness���principle).�6�FP�6sis�justied�b���y�the�informal�principle�of�geometric�completeness,���and��nformally�refutes�Ch���urc�h's��nthesis�CT.�W��*�e�sho���w�that�FP��is�a�constructiv�e���principle�k-in�that�it�is�conserv��q�ativ���e�o�v�er�in�tuitionistic�arithmetic�HA�k'and�has�the���n���umerical���existence�and�disjunction�prop�Gerties.�Returning�to�the�original�philo-���sophical�=Rquestions�ab�Gout�the�nature�of�the�geometric�con���tin�uum,�=Rw�e�ask�what���the���origin�of�our�in���tuitions�is.�W��*�e�sho�w�that�mo�Gdern�ph�ysics�supp�Gorts�the�view�of���Helmholtz���on�this�matter:�b�Gelo���w�the�Planc�k�length�space�is�not�co�Gordinatizable���in�UUthe�usual�w���a�y��*�.��!č�In���tro�s3duction����
src:46ccc.tex�This���pap�Ger�concerns�the�notions�of�existence�and�constructiv���e�existence�as�they���apply�UUto�real�n���um�b�Gers.�UUThe�follo���wing�t�w�o�principles�ha�v�e�often�b�Geen�considered:������
src:48ccc.tex��"V

cmbx10�(Ch��9urc�h's�;thesis)�����':

cmti10�Every��mr��}'e�al�numb�er�c�an�b�e�c�ompute�d�to�any�desir�e�d�ap-���pr��}'oximation���by�an�algorithm.������
src:50ccc.tex�(Geometric���Completeness)�v�The��	p��}'oints�on�a�line�se�gment�c�orr�esp�ond�to���r��}'e�al���numb�ers�in�an�interval.������
src:52ccc.tex�These���t���w�o�statemen�ts�seem�to�b�Ge�dieren�t:�the�rst�one�expresses�an�in�tu-���ition��ab�Gout��p��}'articular��real�n���um�b�ers.��Eac�h�real�n�um�b�Ger�that�someone�migh�t�\giv�e"���to�	jus�is�only�\giv���en"�if�w�e�are�told�ho�w�to�compute�it.�The�second�statemen�t,���on��the�other�hand,�expresses�an�in���tuition�ab�Gout�the��totality��of�real�n�um�b�Gers�(in������|1����*�Q`@�������'`@��������an�^in���terv��q�al).�The�geometric�line�segmen�t�m�ust�b�Ge�\the�sum�of�its�parts"{it�is���comp�Gosed���of�p�oin���ts,�so�there�m�ust�b�Ge�\enough"�p�oin���ts�to�ll�up�the�line,�without���lea���ving�ySan�y�\gaps".�Is�it�really�p�Gossible�that�all�those�gaps�are�lled�up�with���computable�UUn���um�b�Gers?������
src:57ccc.texAll�*5the�notions�in�these�principles�(computation,�algorithm,�real�n���um�b�Ger,���geometric��line�segmen���t)�are�notions�ab�Gout�whic�h�w�e�ha�v�e�strong�in�tuitions,�but���whic���h��.required�Herculean�lab�Gors�of�mathematics�to�bring�to�their�mo�dern�precise���form���ulations.�ҨEv�en�b�Gefore�these�concepts�w�ere�made�precise,�p�Geople�suc�h�as�W��*�eyl���and��bBrou���w�er�w�ere�uncomfortable,�but�with�the�adv�en�t�of�T��*�uring�mac�hines�the���matter��oreac���hed�a�sharp�Ger�form�ulation.�F��*�rom�the�classical�p�Goin�t�of�view,�there�are���only��coun���tably�man�y�computable�real�n�um�b�Gers,�so�no,�not�all�the�gaps�can�b�e���lled��nwith�computable�n���um�b�Gers.��nOf�course,�the�standard�constructivist�reply�is���that�0the�computable�n���um�b�Gers�0are�not�constructiv���ely�en�umerable,�so�y�ou�cannot���p�Goin���t���out�an�unlled�gap.�But�is�this�a�fully�satisfactory�answ�er?�One�migh�t���compare�"�that�answ���er�to�sa�ying,�after�sw�eeping�the�dirt�under�the�carp�Get,�that���y���ou�UUcannot�p�Goin�t�it�out.������
src:59ccc.texIn�UUthis�pap�Ger,�w���e�will�examine�this�question�more�closely��*�.��!č�The�ffcon���tin�uum�in�the�history�of�logic�����
src:63ccc.tex�I�will�b�Gegin,�not�with�the�w���ell-kno�wn�con�tributions�of�Dedekind�and�Can�tor,�but���with�UUanother�historical�stream,�from�the�dev���elopmen�t�UUof�geometry��*�.������
src:65ccc.texGeometry�(�w���as�in�attendance�at�the�birth�of�logic�in�Euclid's��Elements�,�and���the��0nature�of�the�con���tin�uum��0w�as�already�giving�philosophers�dicult�y�b�Gefore�that���(Zeno's���parado���x).�In�the�middle�of�the�nineteen�th�cen�tury��*�,�Staudt�(�Ge��}'ometrie��der���L��}'age�B1847�)�֋to�Gok�steps�to���w�ards�֋a�mo�dern�deductiv���e�geometry;�but�F��*�elix�Klein�ob-���serv���ed�]
in�1873�the�diculties�ab�Gout�con�tin�uit�y�in�Staudt's�treatmen�t,�complaining���of���the�necessit���y�\to�conceiv�e�p�Goin�ts,�also�if�these�are�dened�b�y�means�of�an�in-���nite�[�pro�Gcess,�as�already�existing."�The�second�half�of�the�nineteen���th�cen�tury���sa���w�Mhthe�dev�elopmen�t�of�an�increasingly�rigorous�axiomatic�approac�h�to�geom-���etry��*�,�~�for�example�P���asc�h's�~��V��;�orlesung������ub��}'er��eneuer�e�Ge�ometrie�~��app�Geared�in�1882;���but���as�F��*�reuden���thal�observ�es�in�his�fascinating�history�([�8��],�pp.�106{107),�P�asc�h���had�Hna�n���um�b�Ger�Hnof�Italian�con���temp�oraries:�V��*�eronese,�Enrique,�Pieri,�P���adoa.�These���dev���elopmen�ts��set�the�stage�for�the�app�Gearance�in�1899�of�Hilb�ert's��Grund���lagen���der��7Ge��}'ometrie����[�9��].�����^��ٓ�Rcmr7�1����e�As�F��*�reuden���thal�sa�ys,�the�opinion�is�widespread�that�it�w�as���Hilb�Gert��
who�rst�ga���v�e��
a�completely�deductiv���e�logical�system�for�Euclidean�geom-���etry��*�,��kin�whic���h�nothing�w�as�left�to�in�tuition.�But�in�view�of�the�ac�hiev�emen�ts�of���his�UUpredecessors�just�men���tioned,��what���was�ther��}'e�left�for�Hilb�ert�to�do?����^��2��������
src:72ccc.tex�F��*�reuden���thal��Bansw�ers�this�rhetorical�question�as�follo�ws:�F��*�rom�an�tiquit�y�an���axiom�gw���as�an�eviden�t�truth,�that�could�not�b�Ge�pro�v�ed,�but�also�needed�no�pro�Gof:��X-�ff��D�	J=�����"5��-:��Aa�cmr6�1����L��|{Ycmr8�Quite��p�<rossibly�they�also�in
uenced�the�w�Îork�of�P�eano,�who�in�tro�<rduced�mo�dern�logical��	��notation��Xa�few�y�Îears�later.��	�>�����"5��-:�2����LܻW��J�as��Xblieb�f�<r����ur�Hilb�ert�eigen�Îtlic�h��Xno�c�h��Xzu�tun?�[�8��@],�p.�110.������|�2����;�Q`@�������'`@���������
src:74ccc.tex�Whether���one�b�<reliev�Îed�with�Kan�t�that�axioms�arose�out�of�pure�con�templation,�or����with���Helmholtz�that�they�w�Îere�idealizations�of�exp�<rerience,�or�with�Riemann�that����they�m�w�Îere�h�yp�<rothetical�judgemen�ts�ab�<rout�realit�y��J�,�in�an�y�ev�en�t�nob�<ro�dy�m�doubted����that���axioms�expressed�truths�ab�<rout�the�prop�erties�of�actual�space�and�w�Îere�to�b�e����used��Xfor�the�in�Îv�estigation��Xof�prop�<rerties�of�actual�space.��?G����
src:77ccc.tex�Hilb�Gert's���b�o�ok,�on�the�other�hand,�b�egins�with�\Wir�denk���en�uns�drie�v�er-���sc���hiedene��2Systeme�v�on�Dingen:�die�Dinge�des�ersten�Systems�nennen�wir��Punkte�;���die�s/dingen�des�zw���eiten�Systems�nennen�wir��Ger��}'ade��
�b>

cmmi10�:���:�:�����".�F��*�reuden�thal�sa�ys��\With���this���the�umbilic��}'al�c�or�d�b�etwe�en�r�e�ality�and�ge�ometry�is�sever�e�d."������
src:79ccc.tex�But�oa�few�y���ears�later,�Brou�w�er�w�as�again�attempting�to�c�haracterize�and���elucidate��vthe�prop�Gerties�of�the��actual��con���tin�uum.��vHe�wrestled�for�the�rst�time���with��the�problem�that�will�concern�us�in�this�pap�Ger:�Ho���w�can�w�e�reconcile�the���computabilit���y�t�of�individual�real�n�um�b�Gers�with�our�notion�that�the�con�tin�uum���itself�|is�\full�of�p�Goin���ts",�so�full�as�to�mak�e�a�geometric�line?�Apparen�tly�he���did��Pnot�think�that�one�can�ll�the�gaps�with�computable�n���um�b�Gers:��Phe�in���v�en�ted���his��,theory�of�c���hoice�sequences.�He�said,�for�example,�there�is�a�real�n�um�b�Ger���0.334434333444344��:���:�:�����,��where�w���e�are�free�to�c�ho�Gose�at�an�y�stage�a�3�or�a�4�ar-���bitrarily��*�,�*�according�to�our�free�will.�He�did�not�require�that�w���e�sp�Gecify�at�an�y���nite��stage�an�algorithm�for�making�all�the�rest�of�the�c���hoices.����^��3����&�The�necessit�y���for��Wfunctions�dened�on�[0,1]�to�b�Ge�dened�on�all�c���hoice�sequences�led�to�his���con���tin�uit�y�~yprinciple,�that�all�suc���h�functions�are�con�tin�uous.�This�principle�
atly���con���tradicted�H�classical�mathematics,�and�w�as�resp�Gonsible�in�no�small�part�for�the���w���ell-kno�wn�UUpublic�relations�problems�of�Brou���w�er's�UUin�tuitionistic�mathematics.������
src:84ccc.texHermann�ÁW��*�eyl�w���as�also�concerned�with�the�problem�of�\lling�the�gaps".���W��*�eyl�UU(in�the�preface�to�the�1917�edition�of��Das���Kontinuum�)�wrote:��eӍ��
src:87ccc.tex�A�Ît�uUthe�cen�ter�of�m�y�re
ections�stands�the�conceptual�problem�p�<rosed�b�y�the����con�Îtin�uum{a���problem�whic�Îh�ough�t�to�b�<rear�the�name�of��#�f�cmti8�Pythagor���as��and�whic�h����w�Îe�a�curren�tly�attempt�to�solv�e�b�y�means�of�the�arithmetical�theory�of�irrational����n�Îum�b�<rers.������
src:90ccc.tex�In���the�t���w�en�ties���and�thirties,�Hilb�Gert�applied�his�axiomatic�viewp�oin���t�to���other�Ymathematical�theories�than�geometry��*�,�and�form���ulated�his�program�to�secure���mathematics�A3from�the�dangers�of�the�parado���xes�b�y�pro�ving�the�consistency�of���axiom���systems.�In�his�view,�consisten���t�axioms�had�to�b�Ge�\ab�out�something"{���consistency��guaran���teed�existence.�In�some�sense,�that�is�what�w�as�pro�v�ed�in���the���completeness�theorem�of�G����odel,�but�in�the�t���w�en�ties,���let�alone�the�1910's,���the���concepts�in�the�completeness�theorem�w���ere�not�y�et�clear;�not�un�til�Hilb�Gert-���Ac���k�erman�N�1929�do�Ges�one�nd�the�question�clearly�form���ulated,�and�it�w�as�answ�ered���t���w�o�UUy�ears�later�b�y�G����odel.������
src:92ccc.texIn�UU1932,�W��*�eyl's�b�Go�ok�UUw���as�reprin�ted,�and�he�remark�ed�in�the�preface,���ff��D�	J=�����"5��-:�3����LܻT��J�ec�Îhnically��Kthe�notion�of�c�hoice�sequence�allo�ws�us�to�restrict�our�future�c�hoices;�w�e�could��	��sa�Îy��J�,��}for�example,�that�from�no�w�on�w�e�will�c�ho�<rose�ev�ery�other�digit�to�b�<re�a�3,�and�use�our�free���will��Xonly�on�the�remaining�digits.������|�3������Q`@�������'`@���������
src:94ccc.tex�...in��Ethe�p�<rerio�d��Esince�its�app�<rearance,�m�Îy�w�ork�has�b�<reen�sup�erseded�b�Îy�t�w�o�trends����iden�Îtied��?b�y�the�catc�h�w�ords�In�tuitionism�and�F��J�ormalism.�Still,�this...has�not�led����to�%�an�ev�Îen�mo�<rderately�satisfying�or�defensible�conclusion...it�seems�not�to�b�e����out��_of�the�question�that�the�limitations�prescrib�<red�in�the�presen�Ît�treatise{i.e.,����unrestricted�/�application�of�the�concepts�\existence"�and�\univ�Îersalit�y"�/�to�the����natural�_n�Îum�b�<rers,�but�not�to�sequences�of�natural�n�um�b�<rers{can�once�again�b�e�of����fundamen�Îtal��Xsignicance.�� �H��Recursion�fftheory�and�the�con���tin�uum�����
src:98ccc.tex�The��Qrecursiv���e�con�tin�uum�(if�this�is�not�a�con�tradiction�in�terms)�has�b�Geen�of���in���terest�~�from�the�da�wn�of�computabilit�y�theory:�T��*�uring's�original�pap�Ger�on�T�uring���mac���hines�|had�the�phrase�\computable�n�um�b�Gers"�in�the�title.�A�{�recursiv�e�real���n���um�b�Ger��;is�giv���en�b�y�a�recursiv�e�sequence�of�rational�appro�ximations�con�v�erging���at�UUa�pre-sp�Gecied�rate,�for�example��ak��zy�
!",�

cmsy10�j�x����	0e�rcmmi7�k��$p��8�x����j��6��j����1�=�2�����k���+�8�1�=�2�����j��6��:����src:101ccc.tex�(One�aRcannot�simply�sa���y�that�the�decimal�expansion�is�recursiv�ely�computable���b�Gecause�UUof�tec���hnicalities�ab�out�0.49999��:���:�:�����.)������src:103ccc.texIn�]�the�nineteen-forties,�Stephen�Kleene�dev���elop�Ged�recursiv�e�realizabilit�y��*�,�giv-���ing�Cfor�the�rst�time�a�concrete�and�classically�comprehensible�in���terpretation�of���the��notion�of�\constructiv���e�existence."�The�main�idea�of�recursiv�e�realizabilit�y���is���that�the�quan���tier�com�bination��8�x�9�y�賲is�replaced�b�y�a�recursiv�e�function�that���pro�Gduces�UU�y��.�from��x�.�F��*�or�details�see�[�10��
],�pp.�501-516,�or�[�14��].������src:108ccc.texKleene�UUalso�observ���ed�the�follo�wing�fundamen�tal�fact:��ak����Theorem��T1�(Kleene's�singular�tree)�����B�src:110ccc.tex�K���onig's�K�lemma�is�false�in�the�r��}'e�cursive���c��}'ontinuum.�\�Mor�e�pr�e�cisely,�ther�e�is�an�innite�binary�tr�e�e�with�no�innite�r�e�cur-���sive���p��}'ath.����src:113ccc.texNotation�.�8�W��*�e�use�the�follo���wing�standard�notations�from�recursion�theory:��T�c��(�e;���x;�k�P��)���means�Ithat��k����is�a�computation�b���y�the��e�-th�computable�partial�function�at�input����x�,��}and�the�result�(output)�of�the�computation�is��U��(�k�P��).�W��*�e�also�write��f�e�g�(�x�)�for���this��8v��q�alue��U��(�k�P��).�Since�these�are�partial�functions,��f�e�g�(�x�)�ma���y�b�Ge�undened�for���some����e��and��x�;�w���e�write��f�e�g�(�x�)����T͍��VѸ���+3���VѲ=�����t��y�j�to�mean�that��f�e�g�(�x�)�is�dened�and�is�equal���to�UU�y�[ٲ,�i.e.,��9�k�P��(�T�c��(�e;���x;�k��)�8�^��U��(�k��)��=��y�[ٲ).���ō�src:120ccc.tex�Pr��}'o�of�.�YjOne�starts�with�t���w�o�Yjr.e.,�recursiv���ely�inseparable�sets,�for�example��A�x��=����f�n��;�:��f�n�g�(�n�)����T͍������+3����=������0�g��Ѳand��B��q�f�n��;�:��f�n�g�(�n�)����T͍������+3����=�����1�g�.���The�tree�will�b�Ge�constructed�so�that���an���y��Npath��f�ݲwill�separate��A��and��B��q�:�if��n��a�2��A��N�then��f���(�n�)��a=�0��Nand�if��n��a�2��B�s��then����f���(�n�)�`+=�1.�J�The�denition�of�the�tree��K��is�this:�the�nite�binary�sequence��t��of���length�a�n��b�Gelongs�to��K��}�if�for�eac���h��k������n�,��n��steps�of�computation�of��f�k�P��g�(�k��)�ado�not���rev���eal�UUthat�(�t�)����k�����6�=���f�k�P��g�(�k��);�UUthat�is,����G�%�K�~4�=���f�t��:��:9�j�Y����l�2`h�(�t�)(�T�c��(�k�P�;���k�;�j����)�8�^��U��(�j��)��=�(�t�)����k��됸g�)�:������|�4����+�Q`@�������'`@����������src:124ccc.tex�W��*�e�XBgiv���e�2���^��N��
��the�pro�Gduct�top�ology�and�measure,�induced�b���y�the�norm��j�x�j���=���2���^��O!�cmsy7��k���where��}�k�'�is�the�least�in���teger�suc�h�that��x�(�k�P��)��Z�6�=�0.��}Kleene's�construction�also���sho���ws���that�the�Heine-Borel�theorem�fails�in�recursiv�e�2���^��N�����,�since�the�nite�se-���quences���t��whic���h�are�not�in��K���,�but�whose�initial�segmen�ts�are�all�in��K���,�form�a���co���v�ering�UUof�recursiv���e�2���^��N��
��without�a�nite�sub�Gco�v�er.������src:130ccc.texClosely��related�to�Kleene's�singular�tree�is�the�follo���wing�construction�of�a���\singular�UUco���v�er."��������Theorem��T2�(Lacom��9b�Q�e's�singular�co�v�er)����r9�src:132ccc.tex�The��Yset�of�r��}'e�cursive��Ymemb�ers�of��2���^��N�����has���me��}'asur�e�zer�o.����src:135ccc.texPr��}'o�of�.��?Let����>��0��?b�Ge�giv���en�and�let��k��ֲb�e�a�xed�in���teger�with�1�=�2���^��k�����<���.�Let��y����1��|s�;����:�:�:����;���y����n�����b�Ge���an�en���umeration�of�all�indices�of�partial�recursiv�e�functions��y�齲whose�initial���segmen���ts��6of�length��k��
�+�uv�y��are�dened.�Let��A����n��	!��b�Ge�the�set�of�all�extensions�of�the���initial�\�segmen���t��t��of��f�y����n��q~�g��of�length��y����n���.�+�=��k�P��.�The�measure�of��A����n���
�is�1�=�2���^��y���O
�\cmmi5�n���l�+�k��Dz.�Ev�ery���recursiv���e��mem�b�Ger�of�2���^��N����b�elongs�to�one�of�the��A����n��q~�,�but�their�total�(classically���dened)�UUmeasure�is�b�Gounded�b���y���0���������A�1�������Nϟ�����u

cmex10�X����������n�=1�������<$����1����O�w�feȟ	(֍2���r�y���n���l�+�k��������b������<$��p�1���K�w�fe	둟	(֍2���r�k�������'�<��:�� �l��src:141ccc.tex�R��}'emarks�.�=�Credit�for�this�theorem�is�shared�b���y�Zasla�vski���㎟����i�and��Z��x���Ceitin�[�16��
],�who���made�=
the�argumen���t�constructiv�e�instead�of�classical,�and�adapted�the�construction���to���the�unit�in���terv��q�al,�pa�ying�atten�tion�to�making�the�in�terv��q�als�o�v�erlap�only�at���endp�Goin���ts,�Vbut�Lacom�b�Ge's�publication�w�as�rst.�The�neigh�b�Gorho�o�ds�Vpro�duced�in���Lacom���b�Ge's�Wconstruction�are�not�disjoin�t;�in�general��A����n��ȗ�will�meet��A����m��
ﴲwhen��f�y����n��q~�g����and�UU�f�y����m�����g��ha���v�e�UUa�common�initial�segmen���t�longer�than��max����(�y����n��q~�;���y����m���)�8�+��k�P��.������src:149ccc.texLacom���b�Ge's���construction�is�a�\double�sho�c���k�er":���the�recursiv���e�mem�b�Gers�of���[0,1]�jor�2���^��N��
���ha���v�e�jmeasure�less�than�1,�and�what's�more,�they�ha���v�e�jmeasure�zero!���Con���trary��to�what�one�migh�t�initially�susp�Gect,�neither�of�these�sho�c���ks�is�implicit���in�RIKleene's�construction.�This�matter�is�w���orth�in�v�estigating.�W��*�e�can�get�a�co�v�er���of�d�the�recursiv���e�elemen�ts�of�2���^��N���<�from�Kleene's�tree�b�y�taking�the�collection�of���neigh���b�Gorho�o�ds���determined�b�y�nite�sequences��t��that�do�not�b�Gelong�to�the�tree,���but��)all�their�initial�segmen���ts�do�b�Gelong.�The�elemen�ts�of�this�co�v�er�are�pairwise���disjoin���t.��4What�is�their�total�measure?�One,�or�less�than�one?�The�follo�wing�lemma���answ���ers�UUthis�question.��������Lemma��T1���3��src:153ccc.tex�The��8c��}'over�of�the�set�of�r�e�cursive�memb�ers�of��2���^��N��
#��determine�d�by�Kle�ene's���singular�o�tr��}'e�e�has�total�me�asur�e�1�(classic�al���ly);�c�onstructively,�the�p�artial�sums�of���the���lengths�of�the�c��}'over�ar�e�not�b�ounde�d�by��1�8�������for�any����>��0�.����src:158ccc.texPr��}'o�of�.���Let��U����0��Vb�b�Ge�empt���y�and��K����0�� ��=��2���^��N�����.�Let��U����m�+1�����b�e�the�set�of�mem���b�ers�of�2���^��N�����whose��uinitial�segmen���t�of�length��m����+�1��udo�Ges�not�b�elong�to�Kleene's�tree��K���,�but���whose��Oinitial�segmen���t�of�length��m��do�Ges�b�elong�to��K���.�Then�the��U����m��+�are�pairwise���disjoin���t��:and�their�union�co�v�ers�the�recursiv�e�elemen�ts�of�2���^��N�����.�Let��K����m�+1���زb�Ge�the������|5����;�Q`@�������'`@��������set�Oof�mem���b�Gers�of�2���^��N��
�0�whose�initial�segmen�ts�of�length��m�-4�+�1�Ob�Gelong�to��K���.�Th�us����U����m�+1����[�7J�K����m�+1�����=�C"�K����m�����.��What�sequences��t��of�length��m��are�initial�segmen���ts�of���mem���b�Gers�3�of��U����m�����?�Those�suc�h�that�w�e�ha�v�e��9�j�̂<�9�m�(�T�c��(�j�R;���j�;�m�)��9�^��U��(�m�)�9�=�(�t�)����j��6��).���Giv���en���m�,�there�is�at�most�one��j�dz�suc�h�that��T�c��(�j�R;���j�;�m�);���if�there�is�no�suc���h��j�dz�or�if����j�Y�����m�,��{then��U����m��
_��=��U����m��1����.�Otherwise�(if�there�is�suc���h�a��j����),�then�half�the�sequences���in�V��K����m��1��+��will�drop�out�in��U����m�����.�In�other�w���ords:�for�eac�h��m�,�either��U����m�+1��N�is�empt�y���and�h�K����m�+1����=��G�K����m�����,�or��U����m�+1�� ��consists�of�exactly�half�of��K����m���.�The�total�measure�of���the���U����m��+�is�th���us�(classically)�the�sum�of�the�series�1�=�2�U+�1�=�4�+���:���:�:���V�1�=�2���^��j�����+���:���:�:��
(�=�
z1.���W��*�e��9ha���v�e�not�pro�v�ed�the�constructiv�e�con�v�ergence�of�this�series,�since�w�e�do�not���kno���w��ho�w�large�w�e�m�ust�tak�e��m��to�get�within�a�sp�Gecied�distance�of�1.�But�if�the���co���v�er��had�the�partial�sums�of�its�measures�b�Gounded�b�elo���w�1,�then�there�w�ould�b�Ge���a�&�maxim���um�on�the�v��q�alues�of��m��suc�h�that��U����m�+1����is�nonempt�y��*�,�and�the�tree�w�ould���b�Ge�V�nite,�whic���h�it�is�not.�So�w�e�ha�v�e�pro�v�ed�constructiv�ely�that�for�ev�ery���,�it�is���not�UUthe�case�that�the�partial�sums�are�b�Gounded�b���y�1�8����.������src:175ccc.texIn�b�addition�to�these�failures�in�top�Gology�and�measure�theory��*�,�the�nal�public-���relations�[�disaster�for�recursiv���e�analysis�is�the�failure�in�the�recursiv�e�reals�of�the���theorem���that�b�Gounded�monotone�sequences�ha���v�e���limits.�Sp�ec���k�er���[�13��
]�disco���v�ered���the�Żexistence�of�(what�are�no���w�called)�Sp�Gec�k�er�sequences:�recursiv�e,�strictly�in-���creasing���sequences�of�rationals�b�Gelonging�to�[0,1]�but�not�con���v�erging���to�an���y�re-���cursiv���e���real�n�um�b�Ger.�Suc�h�a�sequence�can�b�Ge�giv�en�as�follo�ws:�W��*�e�construct�the���decimal�u�expansion�of�the�n���um�b�Gers�u�to�con���tain�only�3�and�4�(so�as�to�a�v�oid�prob-���lems�$with�tails�of�nines).�The��k�P��-th�digit�of��x����n�����will�b�Ge�3�if��n��steps�of�computation���of�=8�f�k�P��g�(�k��)�do�not�yield�a�v��q�alue,�or�4�if�they�do�yield�a�v�alue.�This�sequence�is���monotone���since�digits�only�c���hange�from�3�to�4;�and�the�limit�n�um�b�Ger,�if�it�ex-���isted,��Jw���ould�enable�us�to�solv�e�the�halting�problem,�since��f�k�P��g�(�k��)��Jhalts�if�and�only���if�UUthe��k�P��-th�digit�of�the�limit�n���um�b�Ger�UUis�4.������src:180ccc.texThe��Kreisel�basis�theorem�([�12��
],�p.�187)�sa���ys�that�a�recursiv�e�binary�tree���alw���a�ys���has�a����^���0��l�2���	��path.�The�fact�that����^���0��l�2����is�b�Gest-p�ossible���is�illustrated�b���y�the���Kleene's��singular�tree.�K����onig's�lemma�also�fails�in�the�class�of�functions�recursiv���e���in��ӵ�	z�,�for�a�xed����.�T��*�o�a���v�oid���this�phenomenon,�w���e�m�ust�go�to�the�collection�of���partial�o���^���1��l�1�����functions.�K����onig's�lemma�holds�in�this�class:�The�total�functions�in���this��class�are�h���yp�Gerarithmetic,�and�ev�ery�h�yp�Gerarithmetic�innite�binary�tree�has���a�L�h���yp�Gerarithmetic�path.�Th�us�from�a�purely�recursion-theoretic�p�Goin�t�of�view,���there��app�Gears�a�connection�b�et���w�een��the�fullness�of�the�con���tin�uum��and�our�abilit���y���to��dquan���tify�o�v�er�the�in�tegers.�In�a�collection�of�functionals�of�nite�t�yp�Ge,�w�e���need��the�n���umerical�quan�tier��E��5�(considered�as�a�t�yp�Ge�2�functional)�to�guaran�tee���the�>zgeometric�fullness�of�the�t���yp�Ge�1�functions.�(The�h�yp�Gerarithmetic�functions���are��Sexactly�those�functions�of�t���yp�Ge�1�recursiv�e�in��E����.)�F��*�eferman's�analyses�of���predicativit���y�A�[�5��,��6��A�]�sho�w�that�the�h�yp�Gerarithmetic�reals�form�a�mo�del�of�theories���of�0dpredicativit���y��*�,�so�W�eyl�migh���t�consider�his�viewp�Goin�t,�and�the�doubts�quoted���ab�Go���v�e,�UUto�b�e�partially�justied�b���y�these�results.������|6����K��Q`@�������'`@��������Bishop's�ffconstructiv���e�mathematics�����src:191ccc.tex�These��three�failures�of�imp�Gortan���t�classical�results�of�analysis�migh�t�seem�to�b�Ge���the��fdeath�knell�of�Ch���urc�h's��fthesis,�since�they�app�Gear�to�
atly�con���tradict�the���principle��oof�geometric�completeness.�Brou���w�er��odied�in�1966,�so�he�liv���ed�to�see���these�)�results,�but�in�his�usual�st���yle,�he�nev�er�commen�ted�on�them�in�prin�t.�They���ma���y�r�ha�v�e�made�him�glad�that�he�had�dev�elop�Ged�the�theory�of�c�hoice�sequences.���Nev���ertheless,���the�Russian�constructivists�under�the�leader�of�Mark�o�v�pursued���the��dev���elopmen�t�of�constructiv�e�mathematics�assuming�Ch�urc�h's�thesis�for�some���decades.�UUA���t�the�time�of�Brou�w�er's�death�it�app�Geared�that�y�our�c�hoices�w�ere:������src:195ccc.tex(1)��h�ac��}'c�ept��Br�ouwer's�the�ories,�give�up�most�of�mathematics�and�give�up���talking���to�most�mathematicians;�or������src:197ccc.tex�(2)�خ�ac��}'c�ept��CChur�ch's�thesis,�give�up�analysis�and�give�up�talking�to�most���mathematicians;���or������src:199ccc.tex�(3)�UU�r��}'eje�ct���c�onstructive�mathematics�entir�ely.������src:201ccc.tex�This�Fw���as�not�a�dicult�c�hoice�for�most�mathematicians;�but�Errett�Bishop���refused���the�prongs�of�this�dilemma�and�published�a�b�Go�ok���[�2��]�in�1967�(the�y���ear���after�8�Brou���w�er's�death)�in�whic�h�he�dev�elop�Ged�constructiv�e�mathematics�without���using�x�either�Ch���urc�h's�x�thesis�or�c���hoice�sequences.�Since�he�didn't�assume�ev�ery�real���is�Krecursiv���e,�the�recursiv�e�coun�terexamples�do�not�apply�directly��*�.�Since�he�didn't���assume�Hthere�are�some�non-recursiv���e�reals�(e.g.�c�hoice�sequences),�the�classical���theorems�K-are�not�directly�con���tradicted.�His�idea�w�as�to�sho�w�that�b�y�suitable���c���hoices�=of�denitions,�the�constructiv�e�con�ten�t�of�classical�mathematics�could�b�Ge���brough���t�UUto�the�fore,�and�w�as��substantial�.������src:205ccc.texLogicians��ilab�Gored�in�the�subsequen���t�decade�to�analyze�what�Bishop�had���done,�tb���y�constructing�suitable�formal�theories�and�studying�their�formal�in�ter-���pretations.�U�This�w���ork�is�summarized�in�[�1��].�These�studies�v�eried�(for�v��q�arious���formal�|theories)�that�Bishop's�w���ork�is�indeed�consisten�t�with�Ch�urc�h's�thesis�as���w���ell��-as�with�classical�mathematics,�and�is�constructiv�e�in�the�sense�that�\when���a�r/p�Gerson�pro���v�es�r/an�in���teger�to�exist,�he�or�she�can�pro�duce�that�in���teger".�This�is���re
ected�8�in�the�\n���umerical�existence�prop�Gert�y"�of�a�formal�theory�T:�if�T�8�pro�v�es����9�xA�(�x�)�UUthen�for�some�n���umeral���բ����n���	U�,�T�pro�v�es��A�(����M���n�����).��!č�Bishop's�ffmeasure�theory����src:210ccc.tex�One��of�Bishop's�ac���hiev�emen�ts��w�as�the�dev�elopmen�t�of�a�constructiv�e�v�ersion���of�
�measure�theory��*�,�according�to�whic���h�the�unit�in�terv��q�al�has�measure�one.�This���measure���theory�w���as�revised�in�[�3��].�A���similar�revised�v�ersion�app�Gears�in�the�second���edition�N�of�Bishop's�b�Go�ok,�N�whic���h�added�Douglas�Bridges�as�co-author,�and�whose���nal�s-v���ersion�w�as�completed�b�y�Bridges�after�Bishop's�death.����^��4���In�this�section�w�e���extract���from�Bishop's�theory�the�denition�of�\set�of�measure�zero",�and�the���statemen���t�nsand�pro�Gof�of�the�fundamen�tal�lemma�that�p�Germits�Bishop�to�pro�v�e��X-�ff��D�	J=�����"5��-:�4����LܻSee��Xalso�the�discussion�of�the�formalization�of�Bishop's�measure�theory�in�[�7��@].������|�7����]
�Q`@�������'`@��������that�1�sets�of�p�Gositiv���e�measure�are�nonempt�y��*�.����^��5����p�A�t�rst�this�seems�surprising,�since���merely��kno���wing�that�a�set�has�p�Gositiv�e�measure�do�Ges�not�seem�to�pro�vide�enough���information��6to�actually�compute�a�mem���b�Ger�of�the�set.�The�follo�wing�simple�piece���of��<constructiv���e�order�theory�will�b�Ge�needed:�if��u�az�+��v��r<�,��1��<then�either��u�<��1�=�2�or����v�"�<���1�=�2.��One�pro���v�es��this�as�follo���ws:�for�some���>��0�w���e�ha�v�e��u�N\�+��v�"����1�����.��Then�it���is�con���tradictory�that�b�Goth��u�����1�=�2��>+��=�2�and��v�"����1�=�2��>+��=�2.�Hence��u�<��1�=�2��>���=�4���or��v�"�<���1�=�2�����=�4.�A���t�the�last�step�w�e�used�the�usual�constructiv�e�replacemen�t�for���tric���hotom�y:�UUif��a��<�b�UU�then�for�all��x��w���e�ha�v�e��x��<�b�UU�or��a��<�x�.������src:225ccc.texIt���will�turn�out�to�b�Ge�sucien���t�to�understand�the�concept�\set�of�measure���zero"��;in�Bishop's�measure�theory��*�.�Bishop's�measure�theory�applies�to�\comple-���men���ted�UUsets";�for�our�purp�Goses�w�e�can�dene������s�ø�A���=��f�x��:��8�y�"�2��A�(�y��6�=��x�)�g�:����src:227ccc.tex�Here�{all�v��q�ariables�range�o���v�er�{[0,1],�and��x���6�=��y�{T�means�{what�constructivists�usually���call���\apartness";�that�is,�it�means�that�for�some�rational����>��0,����j�x������y�[ٸj���>��.�What���w���e��3usually�call�the�measure�of��A��is�then�the�measure�of�the�complemen�ted�set���(��A;���A�).������src:229ccc.texBishop's�\�denition�on�page�159�of�[�2��]�(with��f�p@�iden���tically�zero)�implies�that����A���has�measure�zero�if�and�only�if�for�eac���h���9X>��0��there�exists�a�sequence�of���nonnegativ���e�UUfunctions��f����j����of�b�Gounded�v��q�ariation�suc�h�that����^��6����|�����src:234ccc.tex�(a)��UU����P�����ލ�
㐷1��%��
㐴j�g��=1�������īR��'��f����j��6��(�x�)����dx�UU�exists�and�is�less�than���.��,����src:236ccc.tex(b)�UU�x���2��A��whenev���er��9��'�>��0�8�N��(�����P�����ލ�
�;�N��%��
�;j�g��=1������f����j��6��(�x�)����1�8����`�).����src:240ccc.texBishop��giv���es�as�an�example�the�case�when��A��is�the�set�of�rational�n�um�b�Gers�in���[0,1].��En���umerate��A��b�y�a�sequence��q����n��q~�.�F��*�or�eac�h����>��0��there�is�a�function��f����n��~:�whic�h���is���1�at��q����n��	"<�and�decreases�linearly�to�zero�on�either�side�of��q����n���so�that�its�in���tegral���is���at�most��=�2���^��n��q~�.�Th���us�condition�(a)�is�satised.�F��*�or�condition�(b):�supp�Gose�the���condition�A�on�the�righ���t�of�(b)�holds�for��x�,�and�let��q����n��	��b�Ge�a�giv�en�rational;�then���since�/�f����n�����decreases�linearly�its�slop�Ge�is�kno���wn,�and�w�e�can�b�Gound��x��a�w�a�y�from��q����n�����in�UUterms�of��n��and���`�,�so��x���6�=��q����n��q~�,�UUso��x���2��A�.������src:247ccc.texThe��#fundamen���tal�lemma�in�Bishop's�measure�theory�(p.�160�of�[�2��],�with��g����iden���tically��E1;�compare�p.�219�of�[�4��])�connects�a�measure-theoretic�statemen�t�with���an�|�existence�statemen���t�ab�Gout�a�p�oin���t.�Essen�tially��*�,�it�sa�ys�that�if�a�set��X�E��has���measure��Fless�than�the�whole�space,�then�w���e�can�nd�a�p�Goin�t��x��in���X���.�W��*�e�will���giv���e���a�more�precise�statemen�t.�In�the�statemen�t,\test�function"�means�con�tin�uous���function�UUwith�compact�supp�Gort.��������Lemma��T2�(Basic�lemma�of�constructiv��9e�measure�theory)���Ы�src:251ccc.tex�L��}'et��صX�_��b�e�a�lo-���c��}'al���ly���c�omp�act�sp�ac�e.�L�et��g�'��b�e�a�nonne�gative�test�function�and�let��f����j����b�e�nonne�gative���|�ff��D�	J=�����"5��-:�5����LܻBishop's�Ydmeasure�theory�is�quite�complicated:�it�has�b�<reen�presen�Îted�in�three�dieren�t�forms��	��in�jthe�literature�and�the�nal�form�in�[�4��@]�o�<rccupies�more�than�eigh�Ît�y�jpages.�It�is�therefore���w�Îorth�while��Xto�extract�here�the�information�relev��an�Ît�to�the�questions�at�hand.��	�>�����"5��-:�6����LܻBishop's��5denition�has�a�condition�(c)�whic�Îh�falls�a�w�a�y�in�the�case�of�measure�zero,�when���the��X��2cmmi8�f���in�Bishop's�denition�is�tak�Îen�to�b�<re�iden�tically�zero.������|�8����	kk�Q`@�������'`@��������test���functions�such�that������P�����ލ���1��%����j�g��=1����V���īR��&e��f����j���T�dx��c��}'onver�ges���and�is�less�than�����īR����g�[ٲ(�x�)����dx�.�Then���we���c��}'an�nd�an��x��in��X�\��and����>��0����such�that�for�al���l�p�ositive�inte�gers��m��`荍������˃�m����������X����t���ly�j�g��=1����;]�f����j��6��(�x�)�����g�[ٲ(�x�)�8���:�� <����src:256ccc.tex�The��basic�idea�of�the�pro�Gof�of�this�lemma�is�considerably�easier�to�grasp�in�a���\totally�ٛdisconnected"�space�suc���h�as�2���^��N�����,�where�the�space�can�b�Ge�divided�in�to,�for���example,��%t���w�o�disjoin�t�subspaces�eac�h�of�measure�half�that�of�the�whole�space.�The���plan�R�of�the�pro�Gof�is�\divide�and�conquer".�Consider�the�illustrativ���e�case�of��X����=��2���^��N�����and�C��g����iden���tically�1.�W��*�e�divide�the�space�in�to�t�w�o�pieces,��A���=��f�f�ڧ�:��f���(0)�=�0�g�C��and����B�G��=���f�f�ڧ�:��f���(0)�=�1�g�.��SGiv���en�a�set�X��5with�measure�less�than�1,�w�e�argue�that�either����X�B��\�y��A�u��or��X��\�y��B��1�has�measure�less�than�1/2.�W��*�e�can�mak���e�this�c�hoice�constructiv�ely��*�,���b���y��the�piece�of�inequalit�y�reasoning�giv�en�ab�Go�v�e:�if��u����+��v�"�<���1��then�either��u�<��1�=�2���or��w�v��)<�IP�1�=�2.�Then�w���e�can�con�tin�ue�in�the�fashion�of�the�usual�pro�Gof�of�Bolzano-���W��*�eierstrass,�~�determining�at�the��n�-th�stage�a�neigh���b�Gorho�o�d�~��U����n��	�<�of�radius�1�=�2���^��n�����suc���h�Ldthat�the�measure�of��X��߸\�&��U����n����is�less�than�half�the�measure�of��U����n��q~�,�and�at�the����n�k^�+�1-st��stage�dividing�the�neigh���b�Gorho�o�d���U����n��	��in�half,�selecting�one�of�the�halv�es���as��G�U����n�+1�����.�Let��g����i��O��b�Ge�c���haracteristic�function�of�the�neigh�b�Gorho�o�d��Gdetermined�at�the����i�-th��stage,�and��x����i��H�its�cen���ter.�Then�the��x����i���form�a�Cauc���h�y��sequence�whose�limit���will���b�Ge�the�desired��x�.�Of�course,�w���e�still�ha�v�e�to�argue�that��x��b�Gelongs�to���X���,�but���this���is�easy:�assume�that������P�����ލ�(�m��%��(j�g��=1���)X�f����j��6��(�x�)�9)�>��1�f������Ʋfor�some�p�Gositiv���e�in�teger��m�.�Then���(b���y�+xthe�con�tin�uit�y�of��f����j��b$�and�the�c�haracteristic�functions��g����i��TL�),�for�sucien�tly�large����n�UU�w���e�ha�v�e�����������q��m������mV)����X����t��m���j�g��=1���}r��f����j��6��(�x�)�g����n��q~�(�x�)�����(1�8����)�g����n���(�x�)�:��i��src:270ccc.tex�In���tegrating,�UUw�e�ha�v�e�����c➟�c��Z��m⟲(1�8����)�g����n��&�dx�����������1����������X����t����j�g��=1���㉵f����j��6��(�x�)�g����n��q~�(�x�)����dx�� <��src:272ccc.tex�con���tradicting�UUthe�h�yp�Gothesis.������src:275ccc.texIf��&the�space�is�not�disconnected�(for�example�[0,1])�then�one�has�to�use�a����p��}'artition��of�unity�.���A���partition�of�unit���y�is�a�nite�collection�of�functions��g����1��|s�;����:�:�:����;���g����m�����whose���sum�is�iden���tically�1,�but�eac�h�of�whic�h�is�nonzero�only�on�a�set�of�small�di-���ameter,��{sa���y�1�=n���^��2��|s�.�In�a�totally�disconnected�space,�partitions�of�unit�y�are�trivial�to���construct;��for�example�in�2���^��N�����,�let��N����t��w�b�Ge�the�set�of�functions�with�nite�initial�seg-���men���t�=��t�,�and�pic�k�an�y�collection�of�neigh�b�Gorho�o�ds�=��N����t����that�co�v�er�the�whole�space,���and�L�let�the��g����i���
�b�Ge�the�c���haracteristic�functions�of�these�neigh�b�Gorho�o�ds.�L�The�use�of���partitions���of�unit���y�is�implicit�in�[�2��],�p.�160-161,�and�more�explicit�in�[�4��],�p.�219,���although�UUin�neither�case�is�the�phrase�\partition�of�unit���y"�actually�men�tioned.������src:284ccc.tex�R��}'emark�.��What�I��w���an�t��to�call�atten���tion�to�is�the�absolutely�crucial�role���pla���y�ed��b�y�the�h�yp�Gothesis�that������P�����ލ�N?�1��%��N?�j�g��=1����O����īR��&��f����j���T�dx��c��}'onver�ges�.�W��*�e�needed�that�n���um�b�Ger������|9����
|��Q`@�������'`@��������to�2�b�Ge�constructiv���ely�w�ell-dened�in�order�to�use�the��u��+��v����argumen�t�2�to�to�decide���whether���to�\go�left�or�righ���t"�at�the��n�-th�stage�in�computing��x�.�If�all�w�e�knew�w�as���that�?�the�partial�sums�w���ere�b�Gounded,�w�e�w�ouldn't�b�Ge�able�to�mak�e�that�decision.��!č�Singular�ffco���v�ers�and�constructiv�e�measure�theory�����src:292ccc.tex�The���main�p�Goin���t�of�this�section�is�to�explain,�without�requiring�a�full�exp�osition���of���constructiv���e�measure�theory��*�,�ho�w�it�is�p�Gossible�that�Bishop's�measure�theory���could��b�Ge�consisten���t�with�Ch�urc�h's�thesis,�in�spite�of�the�singular�co�v�ers�describ�Ged���ab�Go���v�e.��|This�question�is�already�indirectly�approac���hed�in�Exercise�2,�page�70�of�[�1��],���whic���h��addresses�the�follo�wing�tec�hnicalit�y:�the�sum�of�the�lengths�of�the�in�terv��q�als���of�bthe�singular�co���v�er�bis�not�a�recursiv���e�real.�A�Smore�precise�statemen�t�is�that�the���co���v�er�Bconsists�of�a�recursiv���e�sequence�of�in�terv��q�als,�and�the�total�length�of�an�y���nite���n���um�b�Ger�of�them�is�less�than���.�But�the�series�whose�terms�are�the�lengths�is���not�recursiv���ely�con�v�ergen�t.�Putting�the�matter�another�w�a�y��*�,�if�the�singular�co�v�er���consists�j�of�in���terv��q�als��I����n����whose�lengths�are��L�(�I����n��q~�),�then�the�partial�sums��s����n���of�the���series���˟���P��Ȯ�L�(�I����n��q~�)���form�a�Sp�Gec���k�er���sequence.�The�total�length�of�the�co���v�ering���in�terv��q�als���cannot���b�Ge�computed�to�a�predetermined�accuracy��*�.�In�Bishop's�terminology�it���migh���t��^b�Ge�a��fugitive��se��}'quenc�e�{it��^can�alw�a�ys�jump�b�y�an�unkno�wn�amoun�t,�no���matter��`ho���w�long�y�ou�ha�v�e�already�b�Geen�computing.�Exercise�2�asks�the�reader�to���pro���v�e���this�fact,�but�for�a�hin���t�it�suggest�that�otherwise,�Bishop-Cheng�measure���theory�E@w���ould�b�Ge�con�tradicted�if�Ch�urc�h's�thesis�is�assumed,�while�w�e�kno�w�from���metamathematical�"�studies�that�Bishop-Cheng�measure�theory�can�b�Ge�formalized���in��Atheories�that�are�consisten���t�with�Ch�urc�h's�thesis.�This�ma�y�tec�hnically�coun�t���as�L�a�solution,�but�from�the�p�Goin���t�of�view�of�understanding�the�situation,�it�is���circular.�^What�follo���ws�is�a�tec�hnically�useless�attempt�to�pro�v�e�what�w�e�kno�w�to���b�Ge�UUimp�ossible,�but�it�is�helpful�for�understanding.������src:299ccc.texLet�-�us�try�to�use�the�singular�co���v�ers�-�A����n��	�q�dened�ab�Go���v�e�-�to�pro���v�e�-�that�the���set�i��A��of�all�recursiv���e�mem�b�Gers�of�2���^��N��
�>�has�measure�zero,�imitating�the�pro�of�that���the��2rationals�ha���v�e��2measure�zero.�In�2���^��N�����,�c���haracteristic�functions�of�neigh�b�Gorho�o�ds���are�M�con���tin�uous,�w�e�can�tak�e��f����n��	���to�b�Ge�the�c�haracteristic�function�of��A����n��q~�,�whic�h���dep�Gends�u7on�a�giv���en����ev�en�though�the�notation��A����n��浲do�Ges�not�sho�w�this�dep�Gendence.���Condition�#-(b)�w���orks:�if�the�righ�t-hand�side�of�(b)�holds,�then��x��is�not�in�an�y��A����n��q~�,���and�Bqhence�is�not�a�total�recursiv���e�mem�b�Ger�of�2���^��N�����.�T��*�urning�to�(a),�the�in�tegral�������īR��	���f����j��6��(�x�)����dx��is�b�Gounded�ab�o���v�e��b�y�1�=�2���^��y���n���l�+�k��Dz,�where�1�=�2���^��k���1�<�С�,�but�as�w�e�remark�ed���ab�Go���v�e,�Ԙthe�neigh���b�orho�o�ds�in�Lacom���b�e's�co���v�er�Ԙdo�o���v�erlap,�Ԙso�the�actual�(classical)���v��q�alue���of�the�sum�on�the�left�ma���y�b�Ge�less�than�the�b�ound,�and�w���e�do�not�ha�v�e�a���constructiv���e�epro�Gof�that�it�con�v�erges.�The�fact�that�w�e�can't�estimate�the�rate�of���con���v�ergence��eof�this�sum�prev���en�ts��eus�from�pro���ving�that�the�set�of�recursiv�e�reals���has�UUmeasure�zero.������src:307ccc.texLet�i�us�ignore�that�dicult���y�for�a�momen�t,�and�try�to�use�the�fundamen�tal���lemma���of�constructiv���e�measure�theory�to�construct�a�non-recursiv�e�real.�T��*�o�deter-���mine���the�rst�v��q�alue��x�(0)�w���e�consider��U��3�=���f�x��:��x�(0)�=�0�g���and��V����=���f�x��:��x�(0)�=�1�g�.�������10�����ϠQ`@�������'`@��������Let�>��S��J�b�Ge�the�union�of�the�singular�co���v�er,�>�so��S��has�measure��<�L�1�and�w���e�need���to�CVkno���w�whether�it�is��S�k�\�׋�U�Zq�or��S��\�׋�V�|:�that�has�measure�less�than�1�=�2.�Again���w���e��tencoun�ter�the�same�problem:�the�measure�of��S�*�is�not�a�n�um�b�Ger�that�w�e�can���compute�UUto�an���y�desired�accuracy��*�.������src:311ccc.texT��*�o�G�cap�this�discussion,�w���e�will�sho�w�directly�that�the�measure�of�the�union���of�3@the�singular�co���v�er�3@�A����n�����is�not�a�recursiv���e�n�um�b�Ger;�in�other�w�ords,�the�measures���of���the�union�of�the�rst��N���terms�form�a�Sp�Gec���k�er���sequence.�Recall�the�denition���of�
the�singular�co���v�er;�
it�suces�to�tak���e��k���=��1�so�the�measure�comes�out�less�than���1/2.��Then�w���e�en�umerate�as��y����1��|s�;����:�:�:����;���y����n��	�q�those��y�q̲suc�h�that����LщfeB���/��f�y�[ٸg���X�(�y�"�+��I1)�is�dened,���and���w���e�dene��A����n��	qs�to�b�Ge�the�set�of�elemen�ts��f���of�2���^��N�����suc�h�that��f���(�x�)��w=��f�y����n��q~�g�(�x�)���for�D`�x��<�y����n���u�+��1.�Th���us�the�measure�of��A����n���޲is�2���^���y���n���l��1����.�An�y�t�w�o�of�the��A����n���޲are�either���disjoin���t,��ior�one�con�tains�the�other.�Supp�Gose�the�measure�of�the�union�w�ere�a���computable�|�n���um�b�Ger.�Then,�giv�en�a�rational���	>��0,�w���e�could�compute�an�in�teger����K�x�=�N\��(��)��~suc���h�that�for��j���>�K���,�either��A����j���*�is�con���tained�in�one�of�the�rst��K�]��sets��
\i��A����n��q~�,���or�the�measure�of��A����j��ܟ�is�less�than���.�Let��t����j����=��Ms��Lщfe͟�/��f�y����j��6��g���k@�(�y����j���J�+�n�1).�Then�either��f�y�[ٸg����extends�UU�t����n���Ӳfor�some��n�����K���,�UUor�2���^���y���j���a��1���)�<���.������src:322ccc.texThe��Mcondition�2���^���y���j���a��1���)�<����is�equiv��q�alen���t�to��y����j���ĵ>��1���+��lg���~(1�=�),��Mso�it�sa���ys�that�w�e���w���on't�uha�v�e�short�programs��y���coming�out�at�a�late�stage�of�the�en�umeration��y����n��q~�.�W��*�e���will��]sho���w,�ho�w�ev�er,�that�this�p�Gossibilit�y�cannot�b�Ge�prev�en�ted.�The�en�umeration����y����n��_[�is���constructed�in�the�rst�place�b���y�en�umerating�all�computations,�and�putting����y�)��in��the�sequence��y����n��	?��when�w���e�ha�v�e�successfully�computed�all�v��q�alues��f�y�[ٸg�(�x�)�for����x��<�y�[ٲ.��Some�of�those�computations�migh���t�tak�e�a�long�time,�so�a�short�program��y����migh���t���come�out�arbitrarily�late�in�the�sequence��y����n��q~�.�That�is�the�in�tuitiv�e�reason���wh���y��Sthe�measure�of�the�union�of�the�singular�co�v�er�is�not�computable.�W��*�e�can���use��kthe�recursion�theorem�to�mak���e�this�in�tuition�in�to�a�pro�Gof,�as�sho�wn�in�the���next���paragraph.�The�recursion�theorem�p�Germits�us�to�use�the�n���um�b�er���y�LƲin�the���denition�UUof�the�partial�recursiv���e�function��f�y�[ٸg�.������src:332ccc.texBy�]�the�recursion�theorem�dene�a�recursiv���e�function��y��z�as�follo�ws:�to�compute����f�y�[ٸg�(�x�),�s2rst�compute����޲=�2���^���y�@L��1���.�and�s2then��K����=��޵�(��).�Then�compute�(i.e.,�searc���h���for)��La�sequence�n���um�b�Ger��L�t��that�do�es�not�extend��t����n��pʲfor�an���y��n�����K���,��Lbut��t��is�at�least���as�uFlong�as�all�the��t����n���IJwith��n��S���K���.�uFThe�searc���h�for�suc�h�a��t��will�succeed,�since�the���measure�)�of�the�union�of�the�rst��K�࿲of�the��A����n���!�is�less�than�1/2.�Then�the�v��q�alue�to���return���as��f�y�[ٸg�(�x�)�is�(�t�)����x���v�if��x�U(<�l�2`h�(�t�),���and�0�otherwise.�But�b�Gefore�returning�this���v��q�alue,���w���e�(articially)�en�ter�a�long�lo�Gop,�so�that�the�computation�of��f�y�[ٸg�(�x�)�will���tak���e���a�long�time,�sp�Gecically�at�least��K���+�X�1�steps.�Then�the�index��J�{�of��y�ॲin�the���sequence�~h�y����n����will�b�Ge�at�least��K�B&�+��
1,�so�b���y�h�yp�Gothesis�w�e�ha�v�e�either��y��A�extends��t����n����for���some�/��n�����K���,�or�2���^���y�@L��1���<��.�The�rst�alternativ���e�do�Ges�not�hold�since��f�y�[ٸg��extends����t�3w�and��t��is�at�least�as�long�as��t����n�����but�do�Ges�not�extend��t����n��q~�.�Hence�2���^���y�@L��1���<���;�but�b���y���denition�>�of���,�2���^���y�@L��1���=����.�This�con���tradiction�completes�the�pro�Gof.�This�pro�of�giv���es���a��cdirect�solution�of�Exercise�2,�p.�70�of�[�1��],�without�reference�to�Bishop's�measure���theory���or�an���y�metamathematical�argumen�t,�and�thereb�y�demonstrates�wh�y�the���singular�'�co���v�er�do�Ges�not�con�tradict�Bishop's�measure�theory:�the�latter�has�the���h���yp�Gothesis��that�the�sum�of�the�measures�of�the�co�v�er�should�b�Ge�a�constructiv�ely-�������11�����q�Q`@�������'`@��������dened��6real�n���um�b�Ger,��6but�that�h���yp�othesis,�whic���h�is�crucial�for�the�fundamen�tal���theorem��Rthat�sets�of�p�Gositiv���e�measure�con�tain�some�elemen�t,�fails�for�the�singular���co���v�er�UUin�recursiv���e�mathematics.�� ���Geometric�ffcompleteness�and�constructiv���e�logic�����src:353ccc.tex�Bishop's�?�b�Geautiful�construction�of�a�measure�theory�that�is�consisten���t�b�oth�with���classical��Rmathematics�and�with�Ch���urc�h's��Rthesis�is�dazzling,�but�someho���w�the���singular��co���v�er�constructions�still�lea�v�e�one�with�the�feeling�that�Ch�urc�h's�thesis���is�}at�o�Gdds�with�the�geometric�completeness�principle.�In�this�section�w���e�try�to���capture�UUthat�feeling�in�form���ulas.���	����Denition��T1���@sz�src:359ccc.tex�F��;�or�RSsubsets��X�5�of�a�lo��}'c�al���ly�RSc�omp�act�metric�sp�ac�e,�we�dene��F
C�

cmbxti10�X��Uhas���me��i>asur�e��4at�most�t�'��to�me��}'an�that��X���is�c�over�e�d�by�a�union�of�(a�se�quenc�e�of�)���neighb��}'orho�o�ds�/�such�that�the�sum�of�the�me��}'asur�es�/�of�any�nite�numb��}'er�of�those���neighb��}'orho�o�ds���is�less�than�or�e��}'qual�to��t�.������src:365ccc.tex�Th���us�8�the�set�of�recursiv�e�mem�b�Gers�of�[0,1]�has�measure�at�most���,�for�ev�ery������r>��0.�j�This�concept�drops�the�condition�imp�Gosed�b���y�Bishop�that�the�measure�of���the�UUco���v�er�itself�m�ust�b�Ge�computable.������src:369ccc.texWith��ethe�aid�of�this�denition�w���e�can�consider�the�follo�wing�principle,�ex-���pressing�UUthe�\fullness"�of�the�con���tin�uum:������src:372ccc.tex�F��
�ullness��TPrinciple�(FP)�:�UU�If���[0,1]�has�me��}'asur�e���at�most��t��then��t�����1�.������src:374ccc.tex�Although��xFP��*is�mean���t�to�express�an�in�tuition�ab�Gout�[0,1],�its�equiv��q�alen�t���expression��as�a�statemen���t�ab�Gout�2���^��N��|m�is�also�in�teresting.�There�is�a�natural�con-���nection��}b�Get���w�een�binary�trees�and�co�v�ers�of�2���^��N�����.�F��*�or��t��a�nite�binary�sequence�and����f��#�2����2���^��N�����,��:let��t����f��ɲmean�that��f��has��t��for�an�initial�segmen���t.�Eac�h�nite�binary���sequence�P�t��determines�a�neigh���b�Gorho�o�d�P�U����t��k�=�幸f�f��H�:��t����f���g�.�The�co���v�er�asso�Gciated���with���a�tree��T�;�consists�of�all�the��U����t��%�suc���h�that��t�B��62��T��but���all�initial�segmen���ts�of��t����are��Win��T�c��.�Distinct�mem���b�Gers�of�this�co�v�er�are�distinct,�since�if��U����t��k��and��U����s��,��are�in���this��Hco���v�er�then�neither��t��nor��s��has�the�other�for�an�initial�segmen�t.�A��"tree�has����size���at�most��t�F�if�the�sum�of�the�measures�of�an���y�nite�union�of�the��U����t���g�is�at�most����t�.�0�A�0�tree�is��wel���l-founde��}'d��if�ev���ery�path�ev�en�tually�runs�out�of�the�tree.�FP�0�is�then���closely�UUrelated�to�this�statemen���t:������src:386ccc.tex�T��
�ree��EF�ullness�Principle�(TFP)�W��If��Aa�wel���l-founde��}'d�binary�tr�e�e�has�size�at���most���t��then��t�����1�.������src:388ccc.tex�A���t�d�this�p�Goin�t�w�e�review�the�standard�notation�for�sequences�co�Gded�as�in�te-���gers:��w���e�assume�that�ev�ery�in�teger�is�a�sequence�n�um�b�Ger;��l�2`h�(�t�)�is�the�length�of���the��sequence�enco�Gded�b���y��t��and�its�mem�b�Gers�are�(�t�)����0��|s�;����:�:�:����;����(�t�)����n��1����,�where��n���=��l�2`h�(�t�).���W��*�e��write��t�����q�T�or���q�3ܸ��t��to�mean�\�q�T�extends��t�",�that�is,��l�2`h�(�t�)����l�h�(�q�[ٲ))���^�8�j�j�<���l�2`h�(�t�)((�t�)����j��u��=�>�(�q�[ٲ)����j��6��).�6�F��*�or�����a�function�(of�t���yp�Ge�1)�w�e�write��t�>���f�Jb�to�6�mean��8�j��s<���l�2`h�(�t�)((�t�)����j���IJ=���f���(�j����)).�UUW��*�e�write��8��N4�2��2���^��N���	@Y�:���:�:��@U�to�abbreviate��8����(�8�k�P���(�k��)���<��2��!���:���:�:��q��.������src:397ccc.texFP�x�is�y-form���ulated�with�a�function�v��q�ariable�for�the�sequence�of�neigh�b�Gorho�o�ds;���TFP�F�admits�Gsev���eral�formalizations:�it�could�b�Ge�a�second-order�principle�(with�a�set�������12����
�ĠQ`@�������'`@��������v��q�ariable���for�the�tree),�or�a�rst-order�sc���hema,�with�a�form�ula�for�the�complemen�t���of�bthe�tree.�F��*�or�deniteness,�w���e�tak�e�a�v�ersion�that�is�similar�to�FP��*�,�in�that�a���function�UUv��q�ariable���^ϲis�used�to�en���umerate�a�sequence�of�neigh�b�Gorho�o�ds.������src:403ccc.tex������|�8��N4�2���2�����N�����9�k�P��9�n�(���Lщfe{���/�����(�k��)���B�����	z�(�n�))�8�^�8�m������b����������	어m������?�����X����t�����j�g��=0���\p�2������l�
`h�(��(�j�g��))��&0����t������b�����\o�!��t����1��(TFP��� �����src:406ccc.texBy��Ch���urc�h's�thesis�CT,�w�e�mean,�as�usual,�the�assertion�that�ev�ery�se-���quence�-Sof�in���tegers�is�giv�en�b�y�some�recursiv�e�function.�In�view�of�the�singular���co���v�er,�)hFP�)]con�tradicts�Ch�urc�h's�thesis�CT.�Hence�FP�)]is�not�deriv��q�able�in�Bishop's���constructiv���e���mathematics�(BCM)���or�in�formal�systems�that�are�consisten�t�with���CT.������src:412ccc.texNot��only�that,�FP��con���tradicts�\there�exists�a�real������suc�h�that�ev�ery�other���real��is�recursiv���e�in���	z�."�This�can�b�Ge�pro�v�ed�b�y�relativizing�the�singular-co�v�er���construction�UUto�functions�recursiv���e�in���	z�.������src:416ccc.texFP�cis��in���tended�to�express�a�formal�v�ersion�of�the�geometric�completeness���principle,��that�there�are�enough�p�Goin���ts�to�ll�up�a�geometric�line.�Since�it�refutes���CT,�0�it�pro���v�es�0�not-not�there�exists�a�non-recursiv���e�mem�b�Ger���:�of��N����^��N���̲,�so�it�pro�v�es���not-not�X\there�exists�a�non-recursiv���e�subset�of��N�V
��>�N��,�namely�the�graph�of���	z�.�There���is�"�therefore�some�metamathematical�w���ork�to�do�to�c�hec�k�that�this�principle�is���not�UUcompletely�non-constructiv���e.������src:420ccc.texFP���is���carefully�form���ulated�to�a�v�oid�hidden�assertions�ab�Gout�the�computabil-���it���y�,of�n�um�b�Gers;�it�is�in�tended�to�express�that�a�lot�of�p�Goin�ts�exist�(or�tec�hnically��*�,���do��Nnot�fail�to�exist)�to�ll�up�a�line,�without�asserting�that�w���e�can�nd�an�y�sp�Gecic���ones�UUamong�those�p�Goin���ts.�Con�trast�FP�with�the�follo�wing:������src:424ccc.tex�Strong��F��
�ullness�Principle�(SFP)����If���a�subset��X����of�[0,1]�has�me��}'asur�e���at���most���1�8����,�and����>��0�,�then�we�c��}'an�nd�a�memb�er�of���X����.������src:427ccc.tex�This�?|principle�strengthens�the�fundamen���tal�lemma�of�constructiv�e�measure���theory��Hb���y�dropping�the�requiremen�t�that�the�measure�of��X�R*�m�ust�constructiv�ely���exist.������src:430ccc.texW��*�e�K�also�will�consider�sev���eral�constructiv�e�v�ersions�of�K����onig's�lemma,�or�more���precisely��*�,�r�of�\w���eak�K����onig's�lemma".�W�eak�K����onig's�lemma�is�\w���eak"�in�that�the���tree��eis�binary�(or�equiv��q�alen���tly�of�b�Gounded�branc�hing),�rather�than�just�of�nitary���(but���p�Gossibly�un���b�ounded)�branc���hing.�W��*�e�formalize�this�in�theories�without�set���v��q�ariables�I�as�a�sc���hema,�using�a�form�ula�to�represen�t�the�tree.�The�usual�form�ulation���of��\w���eak�K����onig's�lemma�(considered�in�the�pro�Gof�theory�of�subsystems�of�classical���analysis)�UUsa���ys�that�ev�ery�innite�binary�tree�has�an�innite�path.������src:437ccc.texT��*�o�*Lformalize�WKL�*Aand�related�principles,�w���e�m�ust�decide�ho�w�to�represen�t���a���binary�tree:�a�set�v��q�ariable,�or�a�form���ula�(resulting�in�a�sc�hema),�or�a�function���v��q�ariable.�bbT��*�o�c���ho�Gose�the�v�ersion�most�closely�related�to�FP��*�,�w�e�supp�Gose�the�com-���plemen���t��of�the�tree�is�giv�en�b�y�a�sequence�of�neigh�b�Gorho�o�ds.��A���function����~�will���b�Ge��though���t�of�as�en�umerating�sequence�n�um�b�Gers���	z�(�n�),�and�sequence�n�um�b�Ger��t����b�Gelongs��sto�the�complemen���t�of�the�tree�if���	z�(�n�)�����t��s�for�some��n�.�Th�us�the�tree�itself�������13�����"�Q`@�������'`@��������consists��9of�those�sequence�n���um�b�Gers��9�t��suc���h�that��8�n�:�(��	z�(�n�)�Q=���t�).��9In�this�form�ula-���tion,�}-the�closure�of�the�tree�under�subsequence�is�automatic:�if��s�	���t�}-�and��t��do�Ges���not��2extend�an���y���	z�(�n�),�then��s��do�Ges�not�extend�an�y���	z�(�n�)�either,�since�if�it�did,��t����w���ould�UUextend�that�same���	z�(�n�).�W��*�eak�K����onig's�lemma�b�Gecomes:������g���8�m�9�t�(�l�2`h�(�t�)�����m�8�^�8�k�P��:�(��	z�(�k��)�����t�))��!��������g��9��N4�2���2�����N�����8�j����8�k�P��:�(��	z�(�k��)�����\q��3�������������4�(�j����))����(WKL)�������src:446ccc.texConstructiv���ely��*�,�f�w�e�consider�the�v�ersion�\ev�ery�innite�binary�tree�cannot�fail�to���ha���v�e�UUan�innite�path,"�whic���h�w�e�call�\in�tuitionistic�w�eak�K����onig's�lemma":������_�8�m�9�t�(�l�2`h�(�t�)�����m�8�^�8�k�P��:�(��	z�(�k��)�����t�))��!��������_::9��N4�2���2�����N�����8�j����8�k�P��:�(��	z�(�k��)�����\q��3�������������4�(�j����))����(IWKL)�������src:452ccc.texW��*�e�(call�a�tree�\w���ell-founded"�if�ev�ery�path�ev�en�tually�runs�out�of�the�tree.�Th�us���IWKL��sa���ys��1\ev�ery�innite�tree�is�not�w�ell-founded".�The�follo�wing�are�equiv��q�alen�t���expressions��of�IWKL:�\ev���ery�w�ell-founded�binary�tree�is�not�innite,"�and�\there���are�UUno�w���ell-founded�innite�binary�trees."������src:457ccc.texThat��hLacom���b�Ge's�result�(for�some���o�<��1)��halready�implies�Kleene's�(without���assuming�UUthat�all�mem���b�Gers�of�2���^��N��
��are�recursiv�e)�is�the�con�ten�t�of�the�follo�wing:������Theorem��T3���;�D�src:463ccc.tex�With���c��}'onstructive�r�e�asoning,�IWKL�implies�TFP�and�FP.����src:467ccc.texPr��}'o�of�.�ݨSupp�Gose�IWKL;�w���e�will�pro�v�e�TFP��*�.�Supp�Gose�that�2���^��N��
sY�has�measure�at�most����t�.�:�That�is,�there�is�a�co���v�er�:�of�2���^��N�����,�the�partial�sums�of�whose�lengths�are�b�Gounded���b���y�뽵t�.�Since�2���^��N���n�is�totally�disconnected�(op�Gen�neigh�b�Gorho�o�ds��are�also�closed)�w�e���can���dene�a�new�co���v�er���whose�elemen���ts�do�not�o�v�erlap,�b�y�remo�ving�from�eac�h����U����n��	%{�the���part�co���v�ered���b�y�the�union�of�the��U����k�����for��k��q<�d�n�.�Let�the�elemen�ts�of�this���disjoin���t�zlco�v�er�b�Ge�denoted�b�y��V����n��q~�.�Since��V����n��vf���U����n���,�the�partial�sums�of�the�lengths���of�F�V����n��	�q�are�b�Gounded�b���y��t�.�Eac�h��V����n��	�q�is�the�set�of�all��f�mX�2�Yɲ2���^��N��ܤ�whic�h�extend�some���binary�\9sequence��t����n��q~�.�The�prop�Ger�initial�segmen���ts�of�these��t����n��ͷ�form�a�tree��T�c��.�Since����V����n���Ӳis�UUa�co���v�er�UUof�2���^��N�����,�this�tree�has�no�innite�path.������src:477ccc.texHence��$b���y�IWKL,��T��is�not�an�innite�tree.�That�is,�not�not�there�exists�an���in���teger�:v�m��suc�h�that�ev�ery�mem�b�Ger�of��T���has�length�less�than��m�.�If�there�w�ere���suc���h�J9an��m�,�then�since��T��Ȳhas�no�innite�path,�the�sum�of�the�measures�of�the��V����n�����w���ould�.b�Ge�exactly�1.�Hence,�not�not��t�����1.�.But��::�t�����1�.implies��t�����1,�.completing���the�UUpro�Gof.����src:483ccc.tex�R��}'emark�.��zThe�tree��T�Q	�is�not�necessarily�decidable,�since�the�en���umeration�of�the��U����n�����migh���t��
at�an�y�time�spit�out�some�relativ�ely�large�neigh�b�Gorho�o�ds,��
corresp�onding���to��oshort�mem���b�Gers�of��T�c��.�In�tuitiv�ely��*�,��T�[��w�ould�b�Ge�decidable�only�if�the�partial�sums���of�UUthe�measures�of�the��U����n���Ӳdo�not�form�a�\fugitiv���e�sequence".������src:488ccc.texThe�?con���v�erse�question�amoun�ts,�in�tuitiv�ely��*�,�to�whether�the�existence�of���Kleene's��6singular�tree�already�implies�the�existence�of�Lacom���b�Ge's�singular�co�v�er���(dropping���the�assumption�that�all�mem���b�Gers�of�2���^��N�� ��are�recursiv�e).�T��*�ec�hnically�,���it���amoun���ts���to�the�question�whether��:�WKL���implies�not�not�there�exists�a�n�um�b�Ger����t��<��1�UUand�a�singular�co���v�er�UUof�measure�at�most��t�.�������14�����àQ`@�������'`@����������Op�Q�en��TQuestion�1���Z=�src:493ccc.tex�Do��}'es��0FP��imply��I���W�c�K��L�?�Or�even�the�r�estriction�of��I���W�c�K��L����to���de��}'cidable�tr�e�es?����src:496ccc.texR��}'emark�.�Y�W��*�e�sho���w�here�ho�w�one�line�of�attac�k�fails.�Supp�Gose�giv�en�an�innite���binary��tree��T�P!�with�no�innite�path.�W��*�e�m���ust�deriv�e�a�con�tradiction.�F��*�orm�the���co���v�er�xbof�neigh���b�Gorho�o�ds�xb�N����t���ٲ=���f�f��:��t��is�an�initial�segmen���t�of��f���g�,�where��t��is�not�in����T�e��but�all�its�initial�segmen���ts�are�in��T�c��.�(This�step�mak�es�the�extra�assumption���that��o�T�W��is�decidable,�but�ev���en�with�this�assumption�the�argumen�t�w�on't�w�ork.)���By���TFP���the�partial�sums�of�the�measures�of�this�co���v�er���are�not�b�Gounded�b���y�an�y���n���um�b�Ger�%@less�than�1.�This�situation,�ho���w�ev�er,�%@is�not�con���tradictory��*�,�as�sho�wn�in���the��#discussion�of�Kleene's�singular�tree.�In�general,�then,�the�co���v�er��#constructed�in���this�`,w���a�y�from�a�singular�tree�need��not��b�Ge�a�singular�co�v�er.�But�whether�there�is���some���other��w���a�y��to�construct�a�singular�co���v�er��from�a�singular�tree,�I���do�not�kno���w.������src:506ccc.texBrou���w�er's���in�tuition�ab�Gout�the�con�tin�uum�led�him�to�form�ulate�the��fan�.2the��}'o-���r��}'em�,��whic���h�implies�as�a�sp�Gecial�case�that�a�w�ell-founded�binary�tree�is�nite;�this���is�UUessen���tially�Heine-Borel's�theorem�for�2���^��N�����:����B�N�8��N4�2���2�����N�����9�k�P�A�(���\q��l����������/�(�k��))��!�9�m�8��N4�2��2�����N�����9�k�����mA�(���\q��l����������/�(�k��))������src:511ccc.texT��*�o���follo���w�the�pattern�w�e�ha�v�e�used�for�FP���and�IWKL,�w�e�consider�only�the���sp�Gecial�UUcase�in�whic���h�the�predicate��A��is�giv�en�b�y�a�sequence�of�neigh�b�Gorho�o�ds:����ޤ�8��N4�2���2�����N�����9�k�P��9�n�(���\q��l����������/�(�k��)�����	z�(�n�))��!�8��N4�2��2�����N�����9�k�����m�9�n�(���\q��l����������/�(�k��)�����	z�(�n�))��HB�����src:516ccc.texIn���w���ords:�\ev�ery�w�ell-founded�binary�tree�is�nite."�The�con�v�erse�of�HB���is�then���just�x�IWKL,�so�our�in���v�estigations�x�connect�in�this�w���a�y�x�to�the�cen���tury-old�in�v�esti-���gations���of�Brou���w�er;���but�it�seems�to�me�that�an�in���tuition�other�than�the�geometric���completeness���principle�is�needed�to�justify�HB;�and�the�fan�theorem,�whic���h�allo�ws���an���y��;predicate�to�dene�the�complemen�t�of�the�tree,�rather�than�just�a�sequence�of���neigh���b�Gorho�o�ds,��is�(presumably)�ev�en�stronger.�F��*�rom�the�philosophical�viewp�Goin�t���w���e���are�arguing�for�the�acceptance�of�FP���based�on�geometric�completeness,�whic�h���is�UU(on�the�face�of�it)�w���eak�er�UUthan�HB�and�p�Gossibly�w���eak�er�UUthan�IWKL.��!č�Numerical�ffExistence�Prop�s3ert���y�of�FP�����src:526ccc.tex�In� �this�section�w���e�will�use�a�w�ell�kno�wn�form�of�realizabilit�y�to�sho�w�that�FP���has�Ŗthe�disjunction�prop�Gert���y�and�the�n�umerical�existence�prop�Gert�y�when�added���to�zjthe�usual�formal�theories�for�constructiv���e�mathematics.��T�c�F�P���and�zj�W�K��L��can���b�Ge���expressed�in�an���y�of�the�formal�theories�discussed�in�[�1��].�F��*�or�simplicit�y�w�e���consider�w�here�in���tuitionistic�arithmetic�of�nite�t�yp�Ges��H��A���^��!����,�whic�h�has�v��q�ariables�for���in���tegers,��[and�for�functions�of�nite�t�yp�Ge,�and�a�sc�heme�for�denining�function(al)s���b���y��cprimitiv�e�recursion.�WKL��is�formalized�b�y�using�a�t�yp�Ge-1�v��q�ariable�for�the���c���haracteristic���function�of�the�tree.�T��*�o�formalize�TFP�,�a�co���v�er���can�b�Ge�considered���as���determined�b���y�a�sequence�of�neigh�b�Gorho�o�ds,���where�a�neigh�b�Gorho�o�d����N����t��G�is�giv�en�������15�����6�Q`@�������'`@��������b���y�H�the�sequence�n�um�b�Ger��t�.�Th�us�a�co�v�er�is�a�function�from�in�tegers�to�in�tegers.�The���form���ula����f�ڧ�2���t��abbreviates��t����f��L�whic���h�in�term�abbreviates��8�k��<�l�2`h�(�t�)�f���(�k�P��)�=�(�t�)����k��됲,���where��;(�t�)����k��|˲is�the��k��e��`εth��elemen���t�of�the�sequence�co�Gded�b�y�the�in�teger��t�.�Th�us�a���co���v�er�UUof�2���^��N��
��is�a�function���^ϲsuc���h�that���E��~�i�8��N4�2���2�����N�����9�k�P��(���2���	z�(�k�P��))�:����src:538ccc.tex�The�o�measure�of�the�neigh���b�Gorho�o�d�o�t��is�2���^���l�
`h�(�t�)��E��,�so�to�sa�y�that�the�partial�sums�of���co���v�er�UU��^ϲare�b�Gounded�b���y�rational�n�um�b�Ger��r��r�is�to�sa�y�that��2�����̸8�k�������\�k�������?����X����t�����i�=0�����2������l�
`h�(��(�i�))��%N�����r���:������src:541ccc.tex�It�́is�w���ell-kno�wn�́ho�w�to�formalize�the�arithmetic�of�rational�n�um�b�Gers�in�the�arith-���metic�q�of�in���tegers;�the�indexed-sum�functional�can�b�Ge�dened�b�y�the�recursion���sc���heme�UUof��H��A���^��!����.���E����Theorem��T4���;�D�src:544ccc.tex�(i)���H��A���^��!��W�+�O`�F�c�P���has�the�disjunction�pr��}'op�erty���and�the�numeric��}'al�ex-���istenc��}'e���pr�op�erty.������src:546ccc.tex(ii)���If��H��A���^��!��ײ+�8�F�c�P��v�pr��}'oves��9��	zA�(���)�,���then�for�some�term��t�,�it�pr�oves��A�(�t�)�.������src:549ccc.tex(iii)����H��A���^��!��ʩ�+���F�c�P���is�close��}'d�under�Chur�ch's�rule.�Explicitly:�if��H��A���^��!��ʩ�+���F�c�P���pr�oves����8�n�9�mA�(�n;���m�)�,���then�for�some�numer��}'al����!�����e���;��,�it�pr�oves��8�n�9�k�P��(�T�c��(���b:���e����;���n;�k��)�8�^��A�(�n;���U��(�k��))�.����src:554ccc.texR��}'emark�.�_�Since�the�terms�of��H��A���^��!��	)�denote�primitiv���e�recursiv�e�functionals,�(ii)�im-���plies��that��H��A���^��!���+�U�F�c�P��;�is�closed�under�Ch���urc�h's��rule:�if�it�pro���v�es��a�function�with���prop�Gert���y����A��exists,�then�it�pro�v�es�there�is�a�recursiv�e�function�with�prop�Gert�y��A�.���In�	Pparticular,��H��A���^��!��jβ+��׵F�c�P�l߲cannot�pro���v�e�	Pthe�existence�of�a�non-recursiv���e�function,���in�UUspite�of�the�fact�that�it�do�Ges�pro���v�e�UUnot�all�functions�are�recursiv���e.����src:560ccc.tex�Pr��}'o�of�.�?�W��*�e�use�mo�Gdied��q�-realizabilit���y�,�written��e���mq��&��A�.�See�[�14��
],�esp�Gecially�p.���259,�}�Theorem�3.72.�In�view�of�that�theorem�it�suces�to�sho���w�that��H��A���^��!����+�SǵF�c�P����pro���v�es�UUthat��F�c�P���is��mq�-realized.�Let�us�review�the�syn���tactic�form�of�FP:��Y���:�W�8��	z�[�8����9�k�P��(��N4�2�����(�k��))�8�^�8�m�(���������m���������X����t�����i�=0���q�2������l�
`h�(��(�i�))��%N�����t�)��!��t����1]�:��0���src:565ccc.tex�This�phas�the�form��8��	z�[�Q�(��;���t�)��D�!��t����1].�pT��*�o�pro���v�e�pthis�is��mq�-realized:�let����b�Ge���giv���en��and�supp�Gose��j��.��mq��qܵQ�(��	z;���t�)��s�^��Q�(��;���t�).��In�particular��Q�(��;���t�),�so�b���y��F�c�P����w�e�ha�v�e����t�y���1.���The�form���ula��t�y���1���is�a����^���1��l�0���6�form�ula�and�will�b�Ge�realized�b�y�the�iden�tically���zero���function�of�the�correct�t���yp�Ge,�if�true.�Hence��t�����1���is�realized�b�y�the�iden�tically���zero��Gfunction�of�the�t���yp�Ge�to�realize��t��T���1.��GHence���	zz��E��mq���<�F�c�P��.��GThat�completes���the�UUpro�Gof.�� ����ff�Conserv���ativit���y�o�v�er�HA�����src:586ccc.tex�In��Othis�section�w���e�use�in�tuitionistic�forcing�to�pro�v�e�the�conserv��q�ativit�y�o�v�er�HA��*of���some�oRof�the�principles�considered�in�this�pap�Ger.�W��*�e�b�egin�with�some�preliminaries.�������16�����Q`@�������'`@����������src:589ccc.tex�Next,�Y�regarding�the�exact�form���ulation�of��H��A���^��!����:�w�e�can�either�form�ulate��H��A���^��!�����with��.��-terms,�in�whic���h�case�w�e�tak�e��s��to�b�Ge��xy�[�z�p�:xz��(�y�z��),��.or�w���e�can�form�ulate���it�xIusing�com���binatory�logic,�in�whic�h�case��s��is�primitiv�e�and��x:t��denotes�a�term���built��8up�using��k��and��s�.�W��*�e�follo���w�[�14��
]�and�use�the�com�binatory-logic�form�ulation.���W��*�e�4assume�that�the�v��q�ariables�of�t���yp�Ge���r
�are��v���^���[ٴ���n���ԣ�for��n���=�0�;����1�;��2�;��:�:�:��
UO�;�4letters��x�,�y�[ٲ,�and���so�UUon�are�meta-v��q�ariables�ranging�o���v�er�UUthese�actual�ob��8ject�v�ariables.������src:595ccc.texRegarding���notation�for�substitution:�W��*�e�sometimes�write��A�[�x���:=��t�]���to�mean���the��result�of�substituting��t��for��x��in��A�;�more�often�w���e�write��A�(�x�)�and��A�(�t�),�the���former�5�indicating�that��x��ma���y�o�Gccur�free�in��A��and�the�latter�indicating��A�[�x���:=��t�].������src:599ccc.texLet����T��+�b�Ge��H��A���^��!����,�or�some�other�suitable�constructiv���e�formal�theory��*�.�Aug-���men���t�1p�T����b�y�a�constan�t���:�and�an�axiom�that�sa�ys�that���:�denes�a�sequence�of���neigh���b�Gorho�o�ds�UUgiving�the�complemen�t�of�an�innite�binary�tree.��%����_���8�k�P��9�t�(�l�2`h�(�t�)�����k��w�^�8�8�j����:�(��	z�(�j��)�����t�))������src:605ccc.texThat���is,�there�are�arbitrarily�long�sequence�n���um�b�Gers���that�are�not�con���tained�in�an�y���neigh���b�Gorho�o�d�Փ��	z�(�j����).�As�remark�ed�b�Gefore,�the�condition�that�the�set�of�sequence���n���um�b�Gers�{not�con���tained�in�an�y���	z�(�j����)�is�closed�under�subsequence�is�automatic.���This�UUaugmen���ted�theory�w�e�call��T�c��	z�.��t����Lemma��T3���3��src:609ccc.tex�T�c���a�is���c��}'onservative�over��T��.����src:612ccc.texPr��}'o�of�.���Let��Q�(����)�b�Ge�the�axiom�for���	z�,�with�the�constan���t����U�replaced�b�y�a�v��q�ariable�������.��Let��A�(��В�:=�����)�b�Ge�the�result�of�replacing���!\�b���y������in�form�ula��A�.�Dene,�for�eac�h���form���ula�UU�A��of��T�c��	z�,�the�form�ula��A���^�����9�of��T���as�follo�ws:��%��o�B�A���������9�is��e��8����(�Q�(���)���!��A�(��В�:=�����)�;����src:618ccc.tex�where�
p�����is�a�v��q�ariable�that�do�Ges�not�o�ccur�in��A�.�By�induction�on�the�length�of���pro�Gofs��in��T�c�����w���e�ha�v�e:�If��T�c�����pro�v�es��A��then��T����pro�v�es��A���^������.�When��A��do�Ges�not�con�tain�����^ϲthen�UU�A���^�����9�is�pro���v��q�ably�equiv�alen���t�to��A�;�that�completes�the�pro�Gof.������src:623ccc.texIn�UU�T�c��^ϲw���e�can�form�ulate��W�c�K��L��this�w�a�y:��%��F���9����8�k�P�;���j����:�(��	z�(�j��)������\q��3�������������4�(�k��))��HE(WKL��	z�)�����src:627ccc.texIn�͘other�w���ords,�an�y�pro�Gof�in��T�c����plus�this�axiom�can�b�e�syn���tactically�transformed���in���to�UUa�pro�Gof�in��T���from�WKL.������src:630ccc.texW��*�e��will�use�forcing�to�pro���v�e��our�conserv��q�ativ���e�extension�result.�Let��T�c��b��b�Ge���the�_theory��T�c��h��augmen���ted�b�y�another�constan�t��b��for�a�function�from�in�tegers�to���in���tegers.�!yW��*�e�will�dene�forcing�for��T�c��*�in��T��b�.�The�denition�presupp�Goses�that���w���e�,p�Gossess�a�form�ula�of��T�c��	z�,�sa�y��C���(�p�),�that�denes�the�forcing�conditions��p��to���b�Ge�3�used.�Sp�ecically�w���e�dene��p�9l�k��9l�A��for�eac�h�form�ula��A��of��T�c��b�;�the�resulting���form���ula���p���k����A��is�a�form�ula�of��T�c��	z�.�The�atomic�clauses�in�the�denition�of�forcing���are�UUarranged�so�that��%�����õp���k����b�(�n�)�=��m������`�is��������C���(�p�)�8�^��n��<�l�2`h�(�p�)�8�^��(�p�)����n��8��=���m��������:ap���k����A�������is��������A��UU�for��A��atomic�not�in���v�olving�UU�B�����������17�������Q`@�������'`@��������src:640ccc.tex�but�3�since�there�can�b�Ge�atomic�clauses�in���v�olving�3�higher-t�yp�e�3�terms�men���tioning��b�,���this�H�do�Ges�not�suce�as�a�denition.�T��*�o�giv���e�the�correct�denition,�w�e�will�assign���a�*�term���'���N ^���b���t���	�to�eac���h�term��t��of��H��A���^��!����,�of�t�yp�Ge�(1�;����[ٲ)�where����i�is�the�t�yp�Ge�of��t�,�and�then���dene����_e�p���k����t��=��s�m�:=��8����(�p������N4�!���'���ꨲ^���b���t���c���=���-R^����s���w���)�:���J��src:644ccc.tex�The��fdenition�of�the�terms���'����^���b���t���#>�is�as�follo���ws.�In�this�denition,�v��q�ariables��p�,�q�[ٲ,�r�
��are���implicitly�UUrelativized�to�the�form���ula��C�q�dening�the�forcing�conditions.������src:647ccc.tex���э����\q��m�q^������l��b����������:=����������:���������'���o�U�^���b��o�ŵt����������:=����������:t�����for�UU�t��a�constan���t����������i!_^���h��v������[ٴ����k������������:=��������v���1ɍ�[ٴ��;�1���v��k�����������'���l���^���b��j��tq����������:=��������s���'���#��^���b��t�����[�^����r�q��������src:653ccc.tex�so��$that���'�����^���b���tq���L@�(����)��p=�(���'���#�^���b��t����r��)(�����^���q���Ҫ��).��$The�remaining�clauses�in�the�denition�of�forcing,�as���giv���en�UUin�[�1��],�Chapter�XV,�are�as�follo�ws:������B1��p���k���8�xA�����ȏis��������8������^���x�����8�q�"����p�9�r�5���q�[ٲ(�r��k����A�(�x�))��������.��p���k����(�A��!��B��q�)������s:�is��������8�q�"����p�(�q��k����A��!�9�r�5���q�[ٲ(�r��k����B��q�))��������B1��p���k���9�xA������s:is��������9������^���x�����(�p���k����A�)��������:Dp�p���k����A�8�^��B������s:is��������(�p���k����A�)�8�^��(�p��k����B��q�)��������:Dp�p���k����A�8�_��B������s:is��������(�p���k����A�)�8�_��(�p��k����B��q�)��������M/��p���k���?������s:�is�������?�����src:663ccc.tex�The��Pfree�v��q�ariables�of�the�form���ula��p���k����A��are��p�,�together�with�the���f�^����x���
?��suc�h�that��x����is���free�in��A�;�hence�the�use�of���/>^����x����B�instead�of��x��in�the�clauses�ab�Go���v�e���for��9��and��8�.�A���clause�UUfor�negation�is�not�needed�since�w���e�treat��:�A��as��A���!�?�.������src:667ccc.texThe���follo���wing�lemma�is�what�mak�es�forcing�useful�for�conserv��q�ativ�e�extension���results:������Lemma��T4���3��src:670ccc.tex�T�c���a�pr��}'oves���(�p���k����A�)��$��A��for�arithmetic�formulae��A�.����src:673ccc.texPr��}'o�of�.��>A��straigh���tforw�ard�induction�on�the�complexit�y�of��A�,�using�the�lemma�that,���when�UU�A��has�free�v��q�ariables,�(�p���k����A�)[�����^���x���~4�:=���'����^���b���t���c��]��$��p��k��A�[�x��:=��t�].������src:676ccc.texIf�WS���is�a�form���ula�of��T�c��	z�,�then�\���is�generically�v��q�alid"�is�the�form�ula��8�p�9�q�&D�����pq����k��6���.�1�The�soundness�theorem�for�forcing�([�1��],�Ch.�XV,�p.�348),�sa���ys�that�if����T�c��wt�pro���v�es�m���,�and�all�the�axioms�of��T��wt�are�pro���v��q�ably�generically�v�alid�in��T�c��,�then���all�Ctheorems�of��T�c��	��are�pro���v��q�ably�generically�v�alid�in��T�c��.�It�follo���ws�from�Lemma���4��Sthat�if�some�form���ula�or�sc�hema��S�>�is�pro�v��q�ably�generically�v�alid�in��T��then��S�>�is���conserv��q�ativ���e�UUo�v�er��T���for�arithmetic�theorems.������src:683ccc.texAs�UUa�w���arm�up�UUexercise,�w���e�repro�v�e�a�kno�wn�result�of�Simpson:�������Theorem��T5���;�D�src:685ccc.tex�WKL���is�c��}'onservative�over�Pe�ano�arithmetic�P��;�A.��������18����}�Q`@�������'`@��������src:688ccc.tex�Pr��}'o�of�.��
The�tree�dened�b���y�����will�b�Ge�used�to�dene�a�set�of�forcing�conditions,���whic���h�*�w�e�will�use�to�add�a�generic�path��b��through�the�tree.�The�set��C�Ჲof�forcing���conditions�w:is�the�set�of�sequence�n���um�b�Gers�w:�p��that�do�not�ha���v�e�w:only�nitely�man���y���extensions�2�in�the�tree�determined�b���y���	z�.�F��*�ormally�,�2�write��T�c��(�p�)�for��:9�n�(���(�n�)�����p�).���Then�g�T����is�the�tree�whose�complemen���t�consists�of�the�neigh�b�Gorho�o�ds�gen�umerated���b���y�UU��	z�,�and�w�e�dene����@|�C���(�p�)�m:=���T�c��(�p�)�8�^�:9�n�8�q�[ٲ(�q�"���p��^��T�c��(�q�[ٲ)��!��l�2`h�(�q��)����n�)�:����src:695ccc.tex�Since��w���e�are�w�orking�with�classical�logic,�this�is�the�same�as�the�set�of��p��with���innitely�d�man���y�extensions�in��T�c��.�With�classical�logic�then,�w�e�can�pro�v�e�that�for���eac���h�UU�n��there�is�a��p��of�length��n��with��C���(�p�),�from�whic�h�it�follo�ws�that�����6�p���k���8�n�(�T�c��(���\q����������b���c��(�n�))�=�0)�;����src:698ccc.tex�i.e.,���\q���������<��b����h�is�<�a�path�through��T�c��.�F��*�ormalizing�this�argumen���t�w�e�see�that��P�c�A���^��!����b��pro�v�es���that��@�W�c�K��L��is�generically�v��q�alid;�and�as�remark���ed�b�Gefore�the�theorem,�that�is���sucien���t�UUfor�the�conserv��q�ativit�y�of��W�c�K��L��o�v�er��P�c�A��for�arithmetic�theorems.������src:704ccc.texT��*�o���pro���v�e�the�conserv��q�ativit�y�of�IWKL��o�v�er�HA,�w�e�will�not�b�Ge�able�to�im-���itate��Sthe�ab�Go���v�e��Spro�of�directly;�indeed�WKL��5is�not�conserv��q�ativ���e�o�v�er�HA,�so�w�e���cannot���just�add�a�generic�path�through�the�tree��T�]T�determined�b���y���	z�.�Our�pro�Gof�is���more�w�complicated;�forcing�will�b�Ge�com���bined�with�the�mo�del�of��H��A���^��!��	A��in�Kleene's���\coun���table��functionals".�This�notion�is�in�tro�Gduced�in�[�11��
],�and�describ�ed�in�[�1��],���p.�*�135.�W��*�e�review�the�relev��q�an���t�features�of�the�denition�to�establish�notation.���Kleene���denes�the�concept�of�a�t���yp�Ge�1�function�b�eing�an��asso��}'ciate��of�a�t���yp�e������function.�~!This�denition�can�b�Ge�giv���en�in��H��A���^��!��H�b�y�means�of�form�ulas��Ass���^����b��(�f�����^����v%�;���
���8��^��1��
��).���A�UUnotion�of�application�is�dened�on�t���yp�Ge�1�functions�b�y����a2���	z�j��N4�=���n:�(���(�k�P�:��(���\q��l����������/�(�k��))���>��0)�8���1))�;����src:714ccc.tex�where������is�cuto�subtraction.�T��*�erms�of�t���yp�Ge�1�are�dened�to�in�terpret�the�con-���stan���ts����k��and��s��of�t�yp�Ge���X|�as�w�ell�as�the�recursion�constan�ts.�The�functions�of�t�yp�Ge�������will�&Fall�b�Ge�in���terpreted�in�the�mo�del�as�functions�of�t���yp�e�1.����^��7������The�functions�that���will�UUin���terpret�t�yp�Ge����.�are�dened�b�y�a�form�ula��T��c���^�����%�(�
��8�),�giv�en�b�y����d&<�T��c�����(���;���)��w#�(����)��:=��8�x�(�T��c��������%�(�x�)��!��T��c�������bIJ(���j�x�)�:����src:723ccc.tex�Here����x��is�a�t���yp�Ge�1�v��q�ariable.�Of�course�this�starts�with��T��c���^��0����(�x�)��:=��x��=��x����and����Ass���^��0��|s�(��	z;���n�)��:=����(0)�=��n�8�+�1.������src:726ccc.texThe���\mo�Gdel",�expressed�as�a�syn���tactical�in�terpretation,�assigns�to�eac�h�term����t����of��H��A���^��!��o��a�corresp�Gonding�term��t���^����>��and�to�eac���h�form�ula��A��of��H��A���^��!��o��a�corresp�Gonding���form���ula�UU�A���^������,�expressing�that��A��holds�in�the�mo�Gdel,�in�suc�h�a�w�a�y�that��X-�ff��D�	J=�����"5��-:�7����LܻIt�j�is�not�necessary�that�the�in�Îterpretations�of�distinct�t�yp�<res�b�e�disjoin�Ît,�although�this�is�not��	��dicult���to�arrange,�sa�Îy�b�y�using�the�rst�v��alue�of�eac�h�function�as�a�t�yp�<re�lab�el,�and�mo�difying������K�cmsy8�j��>��to��Xignore�(or�t�Îyp�<re-c�hec�k)��Xthose�rst�v��alues.��������19�����Q`@�������'`@����������src:731ccc.tex�(i)�UU(�tq�[ٲ)���^����_��=���t���^������j�q����^���������src:733ccc.tex�(ii)�UU(�v���^���[ٴ��v�k����o�)���^����_��=���v���^���[ٱ1��ٔ�(�k�+B;�@L�)����}�where��v���^���[ٴ��v�k���	IJis�the��k�P��-th�v��q�ariable�of�t���yp�Ge����
x�����src:735ccc.tex�(iii)�UU(�8�x���^����b��A�)���^�����9�is��8����(�T��c���^�����%�(�x���^������)���!��A���^�����)������src:737ccc.tex(iv)�UU(�9�x���^����b��A�)���^�����9�is��9����(�T��c���^�����%�(�x���^������)���!��A���^�����)������src:739ccc.tex(v)�UU�?���^�����9�is��?������src:741ccc.tex�and�%the�map����^�����.�comm���utes�with�the�prop�Gositional�connectiv�es.�Then�the���soundness��#of�the�mo�Gdel�is�expressed�b���y:�if��A��is�a�closed�theorem�of��H��A���^��!��	��then����A���^�������is�a�theorem�of��H��A���^��!����.�T��*�o�pro���v�e�this�w���e�ha�v�e�to�pro�v�e�a�more�general�theorem���applicable�өto�form���ulas��A��with�free�v��q�ariables��x��of�t�yp�Ges������j��
U�in�whic�h�the�conclusion���has�UUthe�v��q�ariables��x���^�����9�relativized�to��T���^���c����;Z�j����%�.������src:747ccc.tex����Sk��Ass�������;���)���(�t;����G�)�8�^��Ass�������b��(�q�[�;�
��8�)���!��Ass��������5�(�tq�[�;��G�j�
��8�)�;����src:748ccc.tex�whic���h�UUis�pro�v��q�able�in��H��A���^��!��	L�for�eac�h�xed�pair�of�t�yp�Ges�(��͠;����!Dz).������src:750ccc.texIn��Jorder�to�use�forcing,�w���e�need�a�v��q�ariation�of�IWKL��%that�asserts�the�ex-���istence���of�something,�so�that�w���e�can�use�forcing�to�add�a�generic�\something".���T��*�o��6construct�suc���h�a�v��q�ariation,�w�e�turn�to�the�G����odel�Dialectica�in�terpretation�of����I���W�c�K��L�o��for�inspiration,�although�the�Dialectica�in���terpretation�is�not�used�in�the���pro�Gof.�t?Namely��*�,�let�us�dene���PathEnder��3ђ�(�e;����	z�)�to�express�that��e��of�t���yp�e�(1,0)�is�a���witness��oto�the�fact�that�the�sequence�����of�neigh���b�Gorho�o�ds��odenes�the�complemen�t���of�UUa�w���ell-founded�tree:����x᭸8�
��8�(��	z�(��e�(�
��)����0����.�)������$�����
������(�e�(�
��8�)����1��|s�))�:����src:756ccc.tex�That��9is,��e�(�
��8�)�is�a�pair�(�n;���k�P��)�suc���h�that�the�initial�segmen�t���"����
���ղ(�k�P��)�of��
�Rq�witnesses�that����
�cl�b�Gelongs��4to�the�neigh���b�orho�o�d�(consisting�of�all�all�extensions�of��)���	z�(�n�).�No���w,�the���form���ula�UUw�e�need�is�the�follo�wing�form�ula�\No�P�ath�Ender"�(NPE)�of��T�c��	z�:����@���9�F�c��8�e�:�(��	z�(��e�(�F��(�e�))����0���%,y�)�������Lщfe@��/��F�c��(�e�)����(�e�(�F�c��(�e�))����1��|s�))��(�N�P�E����)�����src:761ccc.texIn�UUw���ords:��F�c��(�e�)�is�a��
�㍲whic�h�serv�es�as�a�coun�terexample�to���PathEnder��3���(�e;����	z�).������Lemma��T5���3��src:763ccc.tex�H��A���^��!��ײ+�8�AC����1�;�0��l��pr��}'oves���NPE�implies�IWKL.����src:766ccc.texPr��}'o�of�.���Supp�Gose�NPE;�in�order�to�deriv���e�IWKL,�supp�ose���	K�is�a�sequence�of�neigh-���b�Gorho�o�ds�z�dening�the�complemen���t�of�a�w�ell-founded�innite�binary�tree.�W��*�e�m�ust���deriv���e�UUa�con�tradiction.�Since�the�tree�is�w�ell-founded�w�e�ha�v�e����r($�8��N4�2���2�����N�����9�n;���k�P��(���\q��l����������/�(�k��)�����	z�(�n�))�:����src:770ccc.tex�Applying�UU�AC����1�;�0��.;�w���e�ha�v�e�some��e��suc�h�that��e�(����)�is�the�pair�(�n;���k�P��):����m�-�8��N4�2���2�����N�����(���\q��l����������/�(�e�(����)����1��|s�)�����	z�(�e�(���)����0��|s�)�:����src:772ccc.tex�That�(is,���PathEnder��3�e�(�e;����	z�).�But�using�NPE,�w���e�can�construct�a�mem�b�Ger��
�UP�=���F�c��(�e�)���of�$�2���^��N�����suc���h�that��
���is�a�coun�terexample�to���PathEnder��4�5�(�e;����	z�).�This�con�tradiction���completes�UUthe�pro�Gof�of�the�lemma.�������20����-�Q`@�������'`@����������Theorem��T6���;�D�src:777ccc.tex�NPE���is�c��}'onservative�over�HA.����src:780ccc.texPr��}'o�of�.���W��*�e�will�use�forcing�to�add�a�generic�function�(of�t���yp�Ge�1)��f�
�whic�h�will�serv�e���as��Jan�asso�Gciate�of�a�function��F��ٲas�in�NPE.�The�forcing�conditions�will�b�e�sequence���n���um�b�Gers�Iwhic�h�migh�t�b�Ge�initial�segmen�ts�of�suc�h�an�asso�Gciate.�An�asso�ciate�of���suc���h�8Ca�function�w�ould�op�Gerate�on�initial�segmen�ts�of�an�asso�Gciate�of�a�pathender����e����for�the�tree�giv���en�b�y���	z�.�An�initial�segmen�t�of�an�asso�Gciate�of�some�function��e����of�!�t���yp�Ge�(1,0)�can�b�e�visualized�as�a�nite�tree,�the�lea���v�es�!�of�whic���h�are�lab�eled���with���v��q�alues�1�^�+��z��t�where��ݵz���=�%O�e�(����)�for�ev���ery�����extending�that�leaf.�Therefore,�an���initial�F�segmen���t�of�a�pathender��e��for�a�tree�giv�en�b�y�a�sequence�of�neigh�b�Gorho�o�ds������is�
�essen���tially�a�nite�tree�whose�lea�v�es�are�all�co�v�ered�b�y�the�neigh�b�Gorho�o�ds���in���the�range�of���	z�.�Initial�segmen���ts�of��f���m�ust�assign�v��q�alues�to�suc�h�nite�lab�Geled���trees;��@the�v��q�alues�m���ust�b�Ge�zero�if�the�information�in�the�nite�tree�is�not�enough���to�UUdetermine��F�c��(�e�),�or�1�8�+��F��(�e�)�UUif�it�is�enough.������src:791ccc.texA���forcing��condition�will�b�Ge�a�sequence�n���um�b�er��that�could�serv���e�as�an�initial���segmen���t�U�of�an�asso�Gciate��f�l��of��F�c��.�It�th�us�co�Gdes�a�nite�set��X�Ųof�pairs�(�T��V;���v�[ٲ)�where����T�).�is�şa�nite�lab�Geled�tree�and��v�!x�the�asso�ciated�v��q�alue,�and�where�if��v�!x�is�nonzero,���then�b%�v��D��Ak�1�is�a�pair�(�n;���k�P��)�sho���wing�that�no�function��
��]�could�b�Ge�an�asso�ciate�of�a���pathender�S��e��with�an�initial�segmen���t�of��
��Բcorresp�Gonding�to�the�nite�lab�eled�tree����T�c��.�UUSp�Gecically:������src:797ccc.texGiv���en�
a�sequence�n�um�b�Ger��q�h޲sp�ecifying�a�nite�set��X���of�pairs�(�T��V;���v�[ٲ)�where��T����is��^a�nite�lab�Geled�tree,�consider�the�lab�els�on�the�tree�lea���v�es��^as�pairs�(�n;���k�P��)�(ev���ery���n���um�b�Ger�	Mis�a�pair),�and�tak���e�the�maxim�um�v��q�alue��K��i�of�these��k�Y�o�v�er�all�the�trees��T����with�l�(�T��V;���v�[ٲ)�in��X���;�and�tak���e�the�maxim�um�v��q�alue��N���of�these��n�.�The�rst��N��of�the���neigh���b�Gorho�o�ds��T��	z�(�n�)�determine�a�nite�co�v�er;�since����βdetermines�the�complemen�t���of��an�innite�tree,�there�is�a�binary�sequence��t��of�length��max���(�N���;���K���+�[�1)�that�do�Ges���not��extend�an���y���	z�(�n�)�with��n��ø��N��.��An�y�suc�h�sequence��t��will�b�Ge�said�to�b�e�\ok���with�� resp�Gect�to��X���".�The�set��C��<�of�forcing�conditions�is�dened�as�the�set�of��p��suc���h���that,�o�for��q�O_<��l�2`h�(�p�),�if��q��ײsp�Gecies�a�set��X�8�as�ab�o���v�e,�o�then��p����q��^�=��1�J�+��t�o��where��t��is�ok���with��resp�Gect�to��X���;�and�if��q���do�es�not�sp�ecify�suc���h�a�set��X���,�then��p����q����=�,R0.�F��*�orcing���is�UUdened�as�usual,�so�that��p���k����f��(�q�[ٲ)�=�(�p�)����q����for�UUall��q�"�<��l�2`h�(�p�).������src:809ccc.texAs��just�sho���wn,�in��T�c���s�w�e�can�pro�v�e�there�are�arbitrarily�long�forcing�con-���ditions.��W��*�e�no���w�claim�that��N�P�c�E������^����
.��(the�in�tepretation�of�NPE�in�the�Kleene���coun���table-functional���mo�Gdel)�is�generically�v��q�alid.�Supp�ose��p�h1�k��h1��PathEnder���0]S������4�7�(�e;����	z�)���and�#�p��forces���,��determines�an�innite�tree.�W��*�e�m���ust�sho�w�that�some�extension�of����p�[X�forces�a�con���tradiction.�Since��p���k�����PathEnder���0]S������4�7�(�e;����	z�),�if����t�is�an�y�t�yp�Ge-1�function���then�<zev���ery�condition�extending��p��has�a�further�extension�forcing,�for�some��m�,��n�,���and��R�k�P��,��e�p3���(���\q��l����������/�(�m�))�Qh=�1�p3+�(�n;���k��)��Rand���\q��7����������	�n�(�k��)�Qh����	z�(�n�).��RReplacing��m��b���y��max��D�(�m;���k��)�w���e���still�UUha���v�e��e�8���(���\q��l����������/�(�m�))��=�1�8�+�(�n;���k�P��),�and�no���w�w�e�ha�v�e��p���k������\q��l����������/�(�m�)�����	z�(�n�).������src:818ccc.texLet����M���b�Ge�the�length�of��p�,�and�consider�those��q�ŷ<�i�M��suc���h�that��q�زsp�Gecies���a�='nite�set��X�	�of�pairs�(�T��V;���v�[ٲ),�where��T����is�a�nite�lab�Geled�tree.�Then�for�eac���h�suc�h����q�[ٲ,�Pr�p����q��1��=��1�/+��t��where��t��is�ok�with�resp�Gect�to��X���.�Consider,�as�ab�o���v�e,�the�maxim�um���v��q�alues��~�N����and��K�:��of�the��n��and��k���suc���h�that�(�n;���k�P��)�is�a�lab�Gel�on�one�of�the�trees��T����suc���h��4that�(�T��V;���v�[ٲ)�b�Gelongs�to��X���.�As�ab�o���v�e,��4w�e�can�c�ho�Gose�a�sequence��t��of�length�������21����=��Q`@�������'`@��������at��least��max����(�N���;���K�oo�+��S1)�that�do�Ges�not�extend�an���y���	z�(�n�)�with��n�����N��.��Let������b�e���an���y�F}t�yp�Ge�1�function�with�initial�segmen�t��t�.�Cho�Gose��m��and��n��as�ab�o���v�e�so�that����p���k������\q��l����������/�(�m�)�����	z�(�n�).��LThen�in�fact�����(�m�)�������(�n�);��Lin�particular����h�extends����(�n�).�Since����t��u�do�Ges�not�extend�an���y���	z�(�n�)�with��n�����N��,��ubut��t��is�longer�than�all����(�n�)�with��n�����N��,�����}�also��anot�extend�an���y���	z�(�n�)�with��n�����N��.��aThis�is�a�con�tradiction,�and�completes���the��pro�Gof�that�NPE��is�generically�v��q�alid.�Since�for�arithmetic�form���ulae���,�w�e�ha�v�e����p�.��k��.�����^����,L�equiv��q�alen���t��hto���,�the�conserv�ativ���e�extension�result�for��N�P�c�E�&��follo�ws�from���the�UUsoundness�of�forcing.������src:833ccc.texIWKL��0(and��Uhence�FP),�when�added�to��H��A���^��!��bֲ+��ߵAC���,�are�conserv��q�ativ���e�o�v�er���HA.������src:835ccc.texSince�O�IWKL�O�implies�NPE�using��AC����1�;�0��
��,�it�w���ould�suce�to�sho�w�that�if�� ����is���an��Dinstance�of��AC����1�;�0���*�then�� ��[ٟ�^������is�generically�v��q�alid.�That�w���ould�inciden�tally�giv�e���y���et�UUanother�pro�Gof�of�Go�o�dman's�theorem�on�the�conserv��q�ativit���y�of��AC�q�o�v�er�HA.������src:839ccc.texThe�w�in���terpretation�of��AC����1�;�0��P��in�Kleene's�coun�table�functionals�is�equiv��q�alen�t���to�UUthe�follo���wing�principle�of�\con�tin�uous�c�hoice":����8����9�n�(��;���n�)���!�9�g�[ٲ[�8����9�k�P�g��(���\q��l����������/�(�k��))���>��0���^�8�t�(�g�[ٲ(�t�)���>��0��!�8�
�UP���t�(��(�
��8;���g��(�t�)�����1)))]����CC�����src:843ccc.texCC�Xhin�Xiturn�can�b�Ge�decomp�osed:�it�is�equiv��q�alen���t�to�the�conjunction�of��AC����1�;�0��1O�and���\Brou���w�er's�UUprinciple":����4A�8����9�m�(��;���m�)���!�8����9�m;�k�P��8�
��8�(���]����
������(�k��)��=���\q��3������������4�(�k�P��)��!���(�m�))�����B��qP����0��������Theorem��T7���;�D�src:846ccc.tex�H��A���^��!��ײ+�8�C��C�K�is���c��}'onservative�over��H�A�.����src:849ccc.texPr��}'o�of�.��NW��*�e�use�the�tec���hnique�of�[�1��],�Chapter�XV,�namely�,�the�comp�Gosition�of���realizabilit���y���and�forcing.�Let��T�c��b��b�Ge��H��A���^��!�����with�a�constan�t��b��for�a�t�yp�Ge-1�function.���The���realizabilit���y�in�terpretation��e��UU�r��g�A��of��A��go�Ges�from��H��A���^��!��
�Ѳto��T�c��b�.�It�can�b�e���tak���en��Ias�mo�Gdied�realizabilit�y�in�the�coun�table�functionals�recursiv�e�in��b�.�It�is���straigh���tforw�ard�*�to�sho���w�that��C��C���is�realized.�Let����b�Ge�an�arithmetic�sen�tence;���then��in�the�cited�c���hapter�(p.�356)����^��8���
�it�is�sho�wn�ho�w�to�construct�a�notion�of�forcing���suc���h�UUthat����ok��8�p�9�q�"����p�(�q��k����((�e��UU�r��g��)��!���))�:����src:859ccc.tex�Supp�Gose�(T�H��A���^��!���ղ+��޵C��C��p�pro���v�es���.�Then�b�y�the�soundness�of�realizabilit�y��*�,�for�some��e����w���e��aha�v�e��H��A���^��!��hX�pro�v�es��e��UU�r��g��.�By�the�soundness�of�forcing,��H��A���^��!��hX�pro�v�es�that��e��UU�r��g���is���generically�^�v��q�alid.�By�the�prop�Gert���y�of�this�particular�notion�of�forcing,��e��UU�r��g��֖�!������is���pro���v��q�able;�hence��H��A���^��!��
N��pro�v�es���.�But��H��A���^��!��
N��is�conserv��q�ativ�e�o�v�er��H��A�,�so��H�A����pro���v�es�UU��.������Theorem��T8���;�D�src:867ccc.tex�H��A���^��!��0Dz+�fеAC����1�;�0��?��+��I���W�c�K��L��+��F�P�!�is���c��}'onservative�for�arithmetic�the�or�ems���over���H��A�.��ff�ff��D�	J=�����"5��-:�8����LܻIn��-the�cited�reference����is�used�where�w�Îe�no�w�are�using���,�and�on�the�cited�page�there�are��	��t�Îw�o��Xt�yp�<rographical�errors�in�whic�h����is�used�for���.��������22����P��Q`@�������'`@��������src:870ccc.tex�Pr��}'o�of�.���Since�IWKL��implies�FP�in��H��A���^��!����,�and�IWKL�implies�NPE�in��H��A���^��!�����+�ё�AC����1�;�0��
��,���it��suces�to�sho���w�that�NPE��is�conserv��q�ativ�e�for�arithmetic�theorems�o�v�er��H��A���^��!��o��+����C��C��.�f�Supp�Gose�NPE�fapro���v�es�f�an�arithmetic�statemen���t����in��H��A���^��!���
�+���AC����1�;�0��
��.�Then,���using��the�notion�of�forcing�in�the�previous�pro�Gof,��H��A���^��!��^�+�Hg�C��C��4�pro���v�es���8�p�9�q�"����p�(�q��k�����).��Since����is�arithmetic,��q�'�k���N���is�pro���v��q�ably�equiv�alen���t�to���,�b�y�Lemma�4,�so����H��A���^��!��ײ+�8�C��C�q�pro���v�es�UU��.�Then�b���y�the�previous�theorem,��H�A���^��!��	L�pro���v�es�UU��.�� �ߍ�Ph���ysics�ffand�the�Con�tin�uum�����src:879ccc.tex�The��Ytheme�of�this�pap�Ger�is�to�explore�our�geometrical�in���tuitions�ab�out�the�con-���tin���uum.��`In�this�section�w�e�sho�w�that�the�source�of�those�in�tuitions�is�denitely����not���the�nature�of�ph���ysical�space.�Around�1880�the�idea�that�our�geometric�in-���tuitions��w���ere�ab�Gout�ph�ysical�space�w�as�widely�accepted.�Recall�the�quote�from���F��*�reuden���thal�UUgiv�en�earlier�ab�Gout�geometric�axioms:���/���src:884ccc.tex�Whether���one�b�<reliev�Îed�with�Kan�t�that�axioms�arose�out�of�pure�con�templation,�or����with���Helmholtz�that�they�w�Îere�idealizations�of�exp�<rerience,�or�with�Riemann�that����they�m�w�Îere�h�yp�<rothetical�judgemen�ts�ab�<rout�realit�y��J�,�in�an�y�ev�en�t�nob�<ro�dy�m�doubted����that���axioms�expressed�truths�ab�<rout�the�prop�erties�of�actual�space�and�w�Îere�to�b�e����used��Xfor�the�in�Îv�estigation��Xof�prop�<rerties�of�actual�space.������src:887ccc.tex�The�/{dev���elopmen�ts�of�non-Euclidean�and�Riemannian�geometry��*�,�and�their���subsequen���t�a�application�to�the�general�theory�of�relativit�y�b�y�Einstein,�dealt�a���death���blo���w�to�this�idea.����^��9����a�This�to�Gok�place�in�the�early�t�w�en�tieth�cen�tury��*�,�and���sho���w�ed��that�on�v���ery�large�scales�Euclidean�geometry�breaks�do�wn.�Later�it�w�as���also�͌sho���wn�that�Euclidean�geometry�m�ust�break�do�wn�on��smal���l��scales;�ho�w�far���ph���ysics��^has�progressed�to�w�ards�the�utter�destruction�of�Kan�tian�ideas�ab�Gout���space��cdeserv���es�to�b�Ge�more�widely�appreciated�b�y�mathematicians�and�logicians.���What��Hfollo���ws�is�an�explanation�of�the�\Planc�k�length"�and�its�implications�for���the�UUnature�of�space.������src:895ccc.texApparen���tly��>Planc�k�w�as�the�rst�to�note�that�����s0�p���
�?��s0�fep��Ѝ�G�����h���=c���r�3�����,���has�the�dimensions���of�alength,�but�he�oered�no�explanation.�What�follo���ws�is�a�simple�calculation���sho���wing���that�distances�smaller�than�this�length�cannot�exist�in�the�usual�sense;���i.e.,�}�spacetime�cannot�b�Ge�considered�to�b�e�smo�oth�at�that�scale.�The�calcula-���tion��uses�t���w�o��fundamen�tal�equations:�The�uncertain�t�y�principle�from�quan�tum���mec���hanics,�k�and�the�Sc�h�w�arzsc�hild�radius�for�the�formation�of�a�blac�k�hole,�from���general��irelativit���y��*�.�It�is�often�stated�that�\general�relativit�y�and�quan�tum�mec�han-���ics�care�not�consisten���t",�but�seems�not�to�b�Ge�so�w�ell�kno�wn�to�non-ph�ysicists�that���the�Sxinconsistency�can�b�Ge�deriv���ed�in�one�paragraph.�(No�claim�of�originalit�y�is���made�ӽhere;�the�argumen���t�is�w�ell-kno�wn�to�ph�ysicists�and�w�as�sho�wn�to�me�b�y�m�y���friend���Bob�Piccioni.)�These�t���w�o���equations�will�b�Ge�com���bined�to�sho�w�that�there��O��ff��D�	J=�����"5��-:�9����LܻSupp�<rosedly���Gauss�already�attempted�m�Îuc�h���earlier�to�v�Îerify�empirically�that�the�angle�sum��	��of��&a�ph�Îysical�triangle�formed�b�y�moun�tain�tops�is�the�Euclidean�180���-:�q�%cmsy6���UT�.�This�sho�ws�that�he�didn't���think��Xph�Îysical�space�had�to�b�<re�Euclidean.��������23����`ŠQ`@�������'`@��������is��xa�minim���um�radius�giv�en�b�y�the�Planc�k�form�ula�just�men�tioned,�b�Gelo�w�whic�h���spacetime���cannot�b�Ge�regarded�as�smo�oth.�The�smo�othness�of�spacetime�(p�ossi-���bly��except�at�isolated�singularities)�is�a�fundamen���tal�starting�p�Goin�t�for�general���relativit���y��*�,�fLso�this�calculation�sho�ws�the�inconsistency�of�general�relativit�y�and���quan���tum�UUmec�hanics.����^��10��������src:901ccc.tex�The�5�uncertain���t�y�principle�is��E�����t�=R�������h��
R�,�where��E��l�is�the�uncertain���t�y�in���energy���and��t��the�uncertain���t�y���in�time.�Quan���tum�mec�hanics�allo�ws�the�sp�Gon�ta-���neous�#�creation�of�a�particle-an���tiparticle�pair�of�total�mass��M��,�whic�h�can�tra�v�el�a���distance��ǵr�=�and�bac���k�to�annihilate�eac�h�other,�pro�vided�that�the�uncertain�t�y�prin-���ciple��7is�resp�Gected�when�w���e�tak�e�the�uncertain�t�y��E��IJto�b�Ge�the�whole�energy��E����of�'the�particles�and��t��to�b�Ge�their�lifetimes.�Using�Einstein's�equation��E�Z��=���mc���^��2��|s�,���and� 2taking�for�the�lifetime��t��the�time�it�tak���es�ligh�t�to�tra�v�el�the�distance��r�gO�and���bac���k,�UUnamely�2�r�G=c�,�w�e�ha�v�e������/[�M�c�����2��|s�][2�r�G=c�]���������h�����src:908ccc.tex�or��������2�M�r�Gc���������h�����Q0�(1)������src:910ccc.texNo���w�tthe�Sc�h�w�arzsc�hild�solution�of�Einstein's�equations,�expressed�in�units�where����c���=��G��=�1,�UUis���|��Ga$�ds�����2��C��=����(1�8���2�M���=r�G�)�dt�����2���S�+�����<$��NJ�dr����^��2���l�w�fe*?f�	(֍�1�8���2�M�=r�����1��+�8�r������2��Ð� �[ٲ(��;����)��MЍ�src:912ccc.texfor�C�some�function�� ��Ҳof�the�angular�co�Gordinates�����and���.�This�is�v��q�alid�in�the�exterior���of�?�a�spherical�b�Go�dy�?�of�mass��M��.�The�v��q�alue��r���=�M�2�M�Vݲgiv���es�a�zero�denominator;���what�Ĩthis�means�is�that�whenev���er�a�mass�is�compressed�within�its�\Sc�h�w�arzsc�hild���radius"��µr�Ĕ�=�}w2�M��,�the�mass�will�collapse�in���to�a�blac�k�hole.����^��11���;��Putting�the�factors���of�0��c��and��G��in���to�the�equation��r�|I�=�5,2�M��,�w�e�get��r�Gc���^��2�����=�5,2�M�G�.�Solving�for��M�H�w�e���get�M��3�=���r�Gc���^��2��|s�=�(2�G�).�UUPutting�that�in���to�equation�(1)�w�e�ha�v�e������
�r��G����2��Ð�c�����3��|s�=G���������h�����src:922ccc.tex�Solving�UUfor�the�smallest�p�Germissible�v��q�alue�of��r��,�the�Planc���k�length�comes�out:�����4F�r�5��������5�p�������5�fep�		ˍ�����h���G=c���r�3��������src:924ccc.tex�Ev��q�aluating�UUthis�n���umerically�w�e�ha�v�e������Q�r�5����1�:�616�8���10������33�����cm����|�ff��D�	J=�����w���-:�10����LܻThere���are�b�<ro�oks���ab�out�\relativistic�quan�Îtum�mec�hanics",�but�they�are�ab�<rout��sp���e�cial��rela-��	��tivit�Îy��Xand�quan�tum�mec�hanics,�e.g.�the�Dirac�equation.��	�>�����w���-:�11����LܻThe���Sc�Îh�w�arszc�hild�radius�is,�curiously��J�,�what�y�ou�get�if�y�ou�use�the�classical�Newtonian���equation�Q$for�the�escap�<re�v�Îelo�cit�y�Q$�v��I{��-:�2����=�*�2�GM�a>=r����and�set�the�escap�e�v�Îelo�cit�y�Q$equal�to�the�sp�eed���of�YDligh�Ît��c�.�Ho�w�ev�er,�for�establishing�its�connection�to�blac�k�holes,�w�e�need�the�Sc�h�w�arszc�hild���solution��Xof�Einstein's�equations.��������24����p��Q`@�������'`@����������src:927ccc.tex�No���w���w�e'll�go�o�v�er�the�argumen�t�again�without�equations.�The�uncertain�t�y���principle���allo���ws�sp�Gon�taneous�creation�of�sucien�t�energy�to�momen�tarily�(for���Planc���k��@times�on�on�the�order�of�10���^���43��>'�seconds)�collapse�spacetime,�i.e.�creating���a��small�temp�Gorary�blac���k�hole.�Th�us�the�top�Gology�of�spacetime�itself�ma�y�b�Ge���uncertain�r�at�these�dimensions;�it�ma���y�b�Ge�m�ultiply�connected�or�ha�v�e�more�than���three�Espatial�dimensions.�Wheeler�called�this�situation�Quan���tum�F��*�oam.�It�is�w�orth���p�Goin���ting��@out�that�since�the�argumen�t�deriv�es�a�con�tradiction,�w�e�ha�v�en't�really���pro���v�ed�4�the�existence�of�quan���tum�foam�or�of�tin�y�blac�k�holes�at�the�Planc�k�scale.���All��w���e�ha�v�e�pro�v�ed�is�that�something�happ�Gens�at�that�scale�that�w�e�cannot���understand�UUwith�our�presen���t�collection�of�equations�of�ph�ysics.������src:934ccc.texThe��_con���tradiction�b�Get�w�een�quan�tum�mec�hanics�and�relativit�y�implies�that���our�Hin���tuition�of�the�con�tin�uum�do�Ges�not�corresp�ond�to�ph���ysical�realit�y�at�lengths���smaller�.�than�the�Planc���k�length.�The�con�tradiction�dep�Gends�on�general�relativit�y��*�,���and��Ra�fundamen���tal�assumption�of�general�relativit�y�is�that�space�can�b�Ge�assigned���co�Gordinates;�l�or�in�other�w���ords,�n�um�b�Gers�can�b�e�assigned�to�p�oin���ts�on�a�line�in�suc�h���a���w���a�y�that�to�ev�ery�p�Goin�t�there�corresp�Gonds�a�n�um�b�Ger�and�vice-v�ersa.�In�other���w���ords,���the�geometric�completeness�principle�is�assumed�b�y�general�relativit�y��*�.�But���as�*w���e�just�deriv�ed,�this�cannot�con�tin�ue�to�b�Ge�the�case�for�distances�smaller�than���the�UUPlanc���k�length.����^��12���� ����The�ffSource�of�In���tuition�ab�s3out�the�Con�tin�uum�����src:952ccc.tex�Philosophers��ha���v�e�argued�o�v�er�whether�our�in�tuition�of�the�con�tin�uum�is�deriv�ed���from�v6ph���ysical�realit�y��*�,�from�our�exp�Gerience�of�ph�ysical�realit�y��*�,�from�the�nature���of�l\our�minds,�or�from�some�\mathematical�realit���y"�that�w�e�exp�Gerience�with�our���minds.��It�has�no���w�b�Geen�clearly�sho�wn�that�it�is��not��deriv�ed�from�ph�ysical�realit�y��*�.������src:954ccc.texIn�l;the�past�p�Geople�though���t�that�ph�ysical�space�w�as�Euclidean.�Since�Einstein���w���e��hha�v�e�kno�wn�that�our�in�tuition�ab�Gout�lines�do�es�not�corresp�ond�to�the�ph���ysical���\lines"��zdetermined�b���y�ligh�t�paths�in�a�v��q�acuum.�This�w�e�can�call�the�\failure���of���(Euclidean)�geometry�in�the�large."�The�Planc���k-length�argumen�t�w�e�can�call���the�q'\failure�of�geometry�in�the�small."�It�migh���t�b�Ge�argued�that�these�failures���imply��that�our�in���tuition�of�the�con�tin�uum�has�a�non-ph�ysical�source,�whic�h�m�ust���therefore�a�b�Ge�either�our�minds,�or�a�mathematical�but�non-ph���ysical�realit�y�that�can���b�Ge�V�apprehended�b���y�the�mind.�F��*�or�example,�G����odel�suggested�that�our�minds�can���ff��D�	J=�����w���-:�12����LܻThe���Planc�Îk�length�can�b�<re�\disco�v�ered"�in�v��arious�w�a�ys,�the�simplest�of�whic�h�is�to�ask��	��for�]�an�expression�in��G��and��c��that�has�dimension�length.�Another�w�Îa�y�]�is�to�use�the�equation����E���=�/�h����=��hc��P�for�the�energy�of�a�photon�of�frequency���s=�and�w�Îa�v�elength��P��.�This�photon���w�Îould�U�distort�space�in�the�same�w�a�y�as�a�mass�giv�en�b�y��mc���-:�2��\��=�2�E���=��hc=�,�U�so�it�w�ould�ha�v�e���a�Sc�Îh�w�arzsc�hild�radius�of�2�mG=c���-:�2���g�=�~�2�hG=�(�c���-:�3��*���).�No�w�observ�e�that�for����small�enough,������will�Yb�<re�less�than�the�Sc�Îh�w�arzsc�hild�Yradius,�so�the�Sc�Îh�w�arzsc�hild�Ysolution�should�apply�in�the���exterior��nof�the�photon,�and�the�photon�w�Îould�b�<re�sealed�in�to�the�blac�k�hole�of�its�o�wn�creation,���and�g�hence�unobserv��able�b�Îy�us.�The�critical�w�a�v�elength�is�obtained�b�y�setting����equal�to�the��zP�Sc�Îh�w�arzsc�hild��radius�2�hG=c���-:�3��*���.�Solving,�w�Îe�nd���u��=�����೫p���u���೉feޱ�M��2�hG=c�����3�����*T��,��appro�ximately�the�Planc�k���length.�֒This�argumen�Ît,�though�in�teresting�and�strange,�is�not�apparen�tly�con�tradictory��J�,�unlik�e���the��Xargumen�Ît�giv�en�in�the�text.��������25����~�Q`@�������'`@��������serv���e���as�sense�organs�to�apprehend�the�non-ph�ysical�realit�y�of�the�con�tin�uum.�W��*�e���argue��dagainst�this�implication,�instead�supp�Gorting�Helmholtz's�view�(as�quoted���b���y�Z�F��*�reuden�thal)�that�in�tuition�is�deriv�ed�from�our��exp��}'erienc�e��of�ph���ysical�realit�y��*�.���Sp�Gecically��*�,�a�part�of�our�in���tuition�ab�out�the�con���tin�uum�a�is�deriv���ed�as�an�abstraction���from���a�simple�ph���ysical��pr��}'o�c�ess�.���The�pro�Gcess�is�familiar�from�computer�graphics:����zo��}'oming���in�tJ�and��zo�oming���out�.�While�viewing�a�nite�in���terv��q�al�that�is�part�of�a���longer�_line,�w���e�can�double�the�length�of�the�view�ed�in�terv��q�al�(zo�Goming�out)�or���halv���e�:�the�length�of�the�in�terv��q�al�(zo�Goming�in).�Then�w�e�can�adjust�the�size�of�the���displa���y�^#so�that�the�new�selected�in�terv��q�al�app�Gears�congruen�t�to�the�previous�one.���The�^#abstraction�in���v�olv�ed�^#here�is�abstracting�from�the�nite�thic���kness�of�the�line���and��Jthe�small�but�p�Gossibly�nonzero�deviation�from�b�eing�p�erfectly�straigh���t.�After���the���zo�Gom,�the�thic���kness�and�straigh�tness�of�the�(view�of��)�the�line�will�b�Ge�adjusted���to�UUb�Ge�as�b�efore.������src:968ccc.texW��*�e���can�imagine�this�pro�Gcess,�sa���y�zo�oming�in,�going�on�indenitely��*�.�Sp�ecif-���ically��*�,�o�as�man���y�times�as�there�are�p�Gositiv�e�in�tegers.�W��*�e�can�dene��I����0���b�to�b�Ge�[0�;����1]���and��I����n�+1���<�to�b�Ge�the�middle�half�of��I����n��q~�.�The�Planc���k�length�argumen�t�sho�ws�that���this�+#is�not�ph���ysically�the�case{after�ab�Gout�100�zo�oms,�space�itself�is�no�longer���co�Gordinatizable,���and�the�zo�oming-in�pro�cess�breaks�do���wn.�The�exact�manner�of���its�,�failure�is�unkno���wn!�But�nev�ertheless,�the�zo�Goming-in�pro�cess�itself�is�easily���in���tuited,�UUand�w�e�can�distinguish�t�w�o�parts�of�this�in�tuition:��}p�����������src:974ccc.tex�In���tuition�UUof�zo�Goming�in�and�zo�oming�out�(once)��e`�����������src:975ccc.tex�In���tuition�UUof�iterating�a�pro�Gcess�an�y�n�um�b�Ger�of�times������src:978ccc.texThe��in���tuition�of�iterating�a�pro�Gcess�is�reducible�to�the�concept�of�natural���n���um�b�Ger,�UUonce�the�pro�cess�to�b�e�iterated�is�understo�o�d.������src:980ccc.texThis�P�argumen���t�in�supp�Gort�of�Helmholtz's�view�(and�against�G����odel's)�is�not���denitiv���e,���since�the�zo�Goming�pro�cesses�do�not�accoun���t�en�tirely�for�our�in�tuition���of��;the�con���tin�uum.��;Here�are�some�asp�Gects�not�accoun���ted�for:��line��}'arity�,��c�omp�osi-���tion�,�;��c��}'ontinuity�,�and��ful���lness�.�By�linearit���y��*�,�w�e�mean�the�qualit�y�Euclid�had�in���mind���when�he�wrote�that�a�line�is�that�whic���h�has�length�but�not�breadth.�W��*�e���can���zo�Gom�in�on�a�plane�or�ev���en�on�a�self-similar�fractal�set.�By��c��}'omp�osition����w���e���mean�the�question�whether�the�con�tin�uum�is�comp�Gosed�of�(innitely�man�y)����p��}'oints�,�[�eac���h�of�whic�h�has�zero�length,�but�whose�aggregation�can�mak�e�in�terv��q�als���of�jnonzero�length.�The�alternativ���e�conception�is�that�someho�w�these�p�Goin�ts�need���to���b�Ge�activ���ely�created�\at�run�time",�as�a�computer�scien�tist�migh�t�sa�y;�p�Ger-���haps��b���y�Brou�w�er's�\free�c�hoices"�or�b�y�some�sort�of�quan�tum-mec�hanical�device.���The�[�zo�Goming�pro�cesses�also�do�not�address�the��c��}'ontinuity��of�the�con���tin�uum,�[�the���prop�Gert���y�1�that�Dedekind�addressed�with�his�denition�of�completeness�(ev�ery�cut���determines��ta�real)�and�Cauc���h�y��twith�his�denition�of�completeness�(Cauc���h�y��tse-���quences���con���v�erge).�By�sp�Gecifying�that�there�are�no�visible�gaps,�w�e�rule�out�the���p�Gossibilit���y���that�w�e�are�zo�Goming�in�on�some�kind�of�fractal�set�rather�than�the�true���con���tin�uum.���But,�there�are�also�in���visible�gaps�to�w�orry�ab�Gout:�while�zo�oming,�ho���w���can�Atw���e�tell�whether�w�e�are�seeing�the�whole�con�tin�uum�or�only��*�,�sa�y��*�,�the�rational�������26�����;�Q`@�������'`@��������n���um�b�Gers?��DNo�matter�ho���w�man�y�times�w�e�zo�Gom�in�on������P�p���*����P�fe�E���2����*�,�the�gap�in�the�rationals���nev���er�j�b�Gecomes�visible.�W��*�e�can�call�atten�tion�to�it�b�y�placing�a�righ�t�isosceles���triangle���with�its�h���yp�Goten�us���on�the�n���um�b�er���line.�W��*�e�can�similarly�call�atten���tion���to�(�an���y�recursiv�e�real�n�um�b�Ger;�but�ho�w�ab�Gout�the�gap�in�the�recursiv�e�reals�at���the���limit�of�a�Sp�Gec���k�er���sequence?�Can�y���ou�visualize�that�gap?�As�the�predicates���used�C4to�dene�a�Dedekind�cut�increase�more�and�more�in�logical�complexit���y��*�,�the���existence��of�a�p�Goin���t�lling�that�cut�seems�less�and�less�closely�related�to�a�fun-���damen���tal�,kgeometric�in�tuition.�Finally��*�,�the�prop�Gert�y�of��ful���lness��of�the�con�tin�uum,���as�}�axiomatized�in�this�pap�Ger�using�the�fullness�principle�FP��*�,�seems�similar�to���con���tin�uit�y��*�,��Zbut�distinct,�since�the�recursiv���e�reals�satisfy�con�tin�uit�y�(in�the�sense���that�=recursiv���ely�Cauc�h�y�sequences�of�recursiv�e�reals�con�v�erge�to�recursiv�e�reals),���but��they�do�not�satisfy�FP��*�.�F�ullness�do�Ges�not�require�reference�to�sp�ecic�\gaps",���since�Yeit�is�dened�b���y�co�v�erings.�P�erhaps�the�form�ulation�of�the�fullness�prop�Gert�y���and���the�recognition�that�it�is�not�the�same�as�con���tin�uit�y���ma�y�help�in�future�eorts���to�UUelicidate�our�in���tuitions�ab�Gout�the�con�tin�uum.��!č�Conclusions�����src:1004ccc.tex�Our�UUin���tuitions�p�Goin�t�to�t�w�o�principles:������src:1007ccc.tex�(Ch��9urc�h's�;thesis)����Every��mr��}'e�al�numb�er�c�an�b�e�c�ompute�d�to�any�desir�e�d�ap-���pr��}'oximation���by�an�algorithm.������src:1010ccc.tex�(Geometric�n�Completeness)��;�The�A�p��}'oints�on�a�ge�ometric�line�se�gment�c�or-���r��}'esp�ond���to�r��}'e�al���numb�ers�in�an�interval.������src:1013ccc.tex�These�yseem�to�b�Ge�con���tradictory�in�view�of�Kleene's�singular�tree�and�La-���com���b�Ge's��singular�co�v�er.�W��*�e�ha�v�e�formalized�this�feeling�b�y�exhibiting�the�prin-���ciple��FFP��*�,�whic���h�is�justied�b�y�geometric�completeness,�and�con�tradicts�CT.�On���the�UUother�hand�FP�is�otherwise�constructiv���e,�since�������������src:1018ccc.tex�FP�UUhas�the�n���umerical�existence�and�disjunction�prop�Gerties.�������������src:1020ccc.tex�FP��Rsatises��hCh���urc�h's�rule;�in�particular�it�do�Ges�not�pro�v�e�the�existence�of����a�UUnon-recursiv���e�function�������������src:1022ccc.tex�FP��Qis��yconserv��q�ativ���e�for�arithmetic�theorems�when�added�to��H��A���^��!��	�p�or�other����constructiv���e�UUtheories.������src:1025ccc.texThe�8�geometric�con���tin�uum�8�is�\lled"�with�non-recursiv���e�mem�b�Gers,�ev�en�though���w���e���cannot�pro�v�e�their�individual�existence.�P�erhaps�w�e�should�sa�y��*�,�the�con�tin�uum���is�,#not-not�lled�with�non-recursiv���e�mem�b�Gers.�These�unsp�eciable�p�oin���ts�corre-���sp�Gond,�B	p�erhaps,�to�\generic"�reals;�or�p�erhaps,�to�Brou���w�er's�c�hoice�sequences;���or��p�Gerhaps,�some�of�them�can�b�e�generated�b���y�quan�tum-mec�hanical�pro�Gcesses;���or�!Qp�Gerhaps,�they�are�gmen���ts�of�our�mathematical�imagination.�This�conclusion,�������27�����{�Q`@�������'`@��������ho���w�ev�er,���do�Ges�not�necessarily�destro���y�the�basic�premises�of�constructiv�e�mathe-���matics,���nor�do�Ges�it�ev���en�necessitate�accepting�classically�false�axioms�as�Brou�w�er���did.�¨The�principle��F�c�P��,�¨for�example,�is�an�example�of�a�system�expressing�some�of���our�w�in���tuition�ab�Gout�the�non-recursiv�e�"gap-llers"�in�the�con�tin�uum,�and�still�p�Gos-���sessing�tFthe�usual�prop�Gerties�of�constructiv���e�systems.�There�ma�y�b�Ge�other,�stronger���axiom�UUsystems�that�capture�y���et�more�of�our�in�tuition�ab�Gout�the�con�tin�uum.������src:1036ccc.texIn�̱searc���hing�for�suc�h�additional�principles,�it�ma�y�b�Ge�fruitful�to�examine�the���source���of�our�in���tuitions�ab�Gout�the�con�tin�uum.�A�t�an�y�rate,�our�in�tuition�ab�Gout�the���con���tin�uum���is��not��related�to�the�ph���ysical�space�w�e�inhabit,�but�only�to�our�men�tal���conceptions���ab�Gout�a�p�ossible�idealization�of�that�space,�since�mo�dern�ph���ysics�tells���us�UUthat�ph���ysical�space�cannot�b�Ge�co�ordinatizable�and�indenitely�divisible.��!č�References��������[1]����<�src:1047ccc.texBeeson,�@�M.�[1985]��F��;�oundations�C�of�Constructive�Mathematics�.�@�Springer-�����<V��*�erlag,�UUBerlin/�Heidelb�Gerg/�New�Y�ork.������[2]����<�src:1049ccc.texBishop,E.��P[1967]��F��;�oundations�4of�Constructive�A���nalysis�.��PMcGra���w-Hill,�New�����<Y��*�ork.������[3]����<�src:1051ccc.texBishop,�#eE.,�and�Cheng,�H.�[1972]��Constructive�e�Me��}'asur�e�The�ory�.�#eMemoirs�of�����<the�UUA.M.S.��116�.�Pro���vidence,�R.�I.������[4]����<�src:1053ccc.texBishop,��E.,�and�Bridges,�D.�[1985]��Constructive�"�A���nalysis�.�Springer-V��*�erlag,�����<Berlin/�UUHeidelb�Gerg/�New�Y��*�ork�/�T�oky���o�(1985).������[5]����<�src:1055ccc.texF��*�eferman,�/S.�[1964]�Systems�of�predicativ���e�analysis.��J.�8rSymb��}'olic�L�o�gic�/�29�,�����<pp.�UU1{30.������[6]����<�src:1057ccc.texF��*�eferman,���S.�[1968]�Systems�of�predicate�analysis�I�GI.��J.���Symb��}'olic�L�o�gic����33�,�����<pp.�UU193{219.������[7]����<�src:1059ccc.texF��*�eferman,���S.�[1978]�Recursion�theory�and�set�theory:�a�marriage�of�con-�����<v���enience,�^�in�F��*�enstad,��et.��*al.�,��Gener��}'alize�d�R�e�cursion�The�ory�II�,�^�pp.�55{98,�����<North-Holland.������[8]����<�src:1062ccc.texF��*�reuden���thal,�~�H.�[1957]�Zur�Gesc�hic�h�te�der�Grundlagen�der�Geometrie.�Nieu�w�����<Arc���hief�UUv�o�Gor�Wiskunde,�derde�serie,�deel�V,�No.�3,�pp.�105{142.������[9]����<�src:1064ccc.texHilb�Gert,��sD.�[1899]��Grund���lagen���der�Ge��}'ometrie�.��sB.�G.�T��*�eubner,�Stuttgart.�The�����<ten���th��edition�(1968)�w�as�translated�in�to�English�and�published�as��F��;�ounda-�����<tions���of�Ge��}'ometry�,�UUOp�Gen�Court,�La�Salle,�Illinois�(1971).�����[10]����<�src:1067ccc.texKleene,�;'S.�[1952]��Intr��}'o�duction��Nto�Metamathematics�.�;'v��q�an�Nostrand,�Princeton.�������28�������Q`@�������'`@����������[11]����<�src:1069ccc.texKleene,��S.�[1959]�Coun���table�functionals.�in:��Constructivity��in�Mathematics�����<(Pr��}'o�c�e�e�dings�of�the�Col���lo��}'quium�at�A���mster�dam,�1957)�.���Edited�b���y�A.�Heyting.�����<North-Holland,�UUAmsterdam.�����[12]����<�src:1072ccc.texSho�Geneld,�ezJ.�[1967]��Intr��}'o�duction��Fto�Mathematic��}'al�L�o�gic�.�ezAddison-W��*�esley�,�����<Reading,�UUMass.�����[13]����<�src:1074ccc.texSp�Gec���k�er,�
�E.�[1949]�Nic���h�t�
�k�onstruktiv�b�Gew�eisbare�S����atze�der�Analysis.�Journal�����<of�UUSym���b�Golic�Logic��14�,�pp.�145-148.�����[14]����<�src:1076ccc.texT��*�ro�Gelstra,�A.�S.�[1973]��Metamathematic��}'al��Investigation�of�Intuitionistic�����<A���rithmetic��and�A�nalysis�.���Lecture�Notes�in�Mathematics��344�.�Springer,�����<Berlin.�����[15]����<�src:1077ccc.texW��*�eyl,���H.�[1918]��Das�A�Kontinuum:�Kritische�Untersuchunger�bT����ub��}'er�die�Grund-�����<lagen��Wder�A���nalysis�,�j�V��*�eit,�Leipzig.�English�translation:��The��WContinuum:�A�����<Critic��}'al�~FExamination�of�the�F��;�oundation�of�A���nalysis�,�=�Do���v�er,�Mineola,�N.�Y.,�����<1994.�����[16]����<�src:1081ccc.texZasla���vski���㎟����i,��+I.�D.,�and������x���Ceitin,�G.�S.�[1962]�Singular�co�v�erings�and�prop�Gerties�����<of���constructiv���e�functions�connected�with�them.�T��*�rudy�Mat.�Inst.�Steklo�v��67�,�����<1962,��5pp.�458{502.�English�translation�in:�A.M.S.�T��*�ranslations��(2)�.p98�,�1971,�����<pp.�UU41-89.�������29���������;�Q`@]�		�F
C�

cmbxti10�#�f�cmti8��':

cmti10��"V

cmbx10���N�ffcmbx12�q�%cmsy6��K�cmsy8��2cmmi8��Aa�cmr6�|{Ycmr8�X�Qcmr12�D��tG�G�cmr17�
!",�

cmsy10�O!�cmsy7�
�b>

cmmi10�	0e�rcmmi7�O
�\cmmi5�K�`y

cmr10�ٓ�Rcmr7���u

cmex10�������

Sindbad File Manager Version 1.0, Coded By Sindbad EG ~ The Terrorists