HIERARCHIES OF SETS AND DEGREES BELOW 0' by Richard L. Epstein, We examine of changes two hierarchies a recursive generalizations by asking what [4]. [8]. ordinal, [6]. hierarchies of sets below 0' based on the number approximation functions error predicates by Ershov and Richard Lo Kramer* to a set needs of the notion of being r.e. set out in Epstein constructive Richard Haas, dominate as previously The second views the changes as first suggested by Addison a translation between to the degrees of u n s o l v a b i l i t y as dominated by a [i], and developed them and relate these i 0'. All n o t a t i o n < 0' which is based on the jump operator to Cooper. See Epstein We indicate by comes [4]. The reader should be aware that there is another hierarchy degrees sets the ideas of Putnam on trial and We first review some facts about sets iT 0' from Epstein Both are The first classifies the number of changes, This extends We provide to make. of and is due primarily [4], Chapter XI for that. the end of a proof by a , and the end of a subproof D. *We are grateful to H. Hodes an earlier version of this paper. suggestions concerning diagrams. for correcting a number of errors Roger Maddux also aided us w i t h in We notate 0' = {(x,y): ~ (y)+}. Recall that 0' ~T K = {x : ~ x ( X ) + } . First we note the Q u a n t i f i e r C h a r a c t e r i z a t i o n of Sets Below 0': A ~T 0' iff there are two r e c u r s i v e predicates that Proof: =_J w E A Vy S(x,y,w) Therefore Similarly, is r.e. in 0' so A is r e c u r s i v e in =_J some Y ! s} Let f Vx~y R(x,y,w) iff XxVy S(x,y,w) is recursive ment of an r.e. predicate. ~ x V y NR(x,y,w) iff such in 0' since its the comple- ~ x V w S(x,y,w) in 0' R,S Thus is r.e. A and X in 0' are r.e. 0' enumerate 0' and let be 0' e n u m e r a t e d to level s. 0's = {x : f(y) = x _< s, For some e, A = ~e(0') the e th f u n c t i o n partial r e c u r s i v e in 0'. Define As(X) w i t h calculations truncated by = ~e,s(0~)(x) s). (where +e,s is e Then x E A iff ~tVs > t ~e,t(0')(x) = 1 iff Vt~s > t ~e,t(0')(x) = 1 m As a c o r o l l a r y to the proof we have The Limit L e m m a sive function (Shoenfield g [II]): f iT 0' iff there is some recur- such that f(x) = lim s g(s,x)- Here lim s g(s,x) = f(x) means for all sufficiently large g(s,x) ~tVs ~ t = a = f(x). That is, W h e n we have g(s,x) g(s,x) = a. as in the Limit Lemma we call g(s,x) = fs(X) and call that a r e c u r s i v e a p p r o x i m a t i o n to we'll v i e w it as a set and say If As(X) s (if A is r.e. "changes (if x E A x ~ A, then A there is one) The r.e. g(s,x) = 0 f. If such that f is 0-i v a l u e d = As(X). has an a p p r o x i m a t i o n its m i n d at most once." As(X) s, As such that That is, there is at most one As(X) = 0 and As+l(X) = 1 always). sets have always been c o n s i d e r e d a d i s t i n g u i s h e d class of 34 sets b e l o w 0'. This is partly because the special p r o p e r t y just d e s c r i b e d is easy to utilize, tion in logic between axioms set), as "proving" and p a r t l y because it reflects the rela(a r e c u r s i v e set) and theorems (an r.e. is just a recursive enumeration. We w i s h to generalize the notion of r.e.'ness so that r.e. sets are seen as part of a continuum. Let us classify A by m e a s u r i n g how often an a p p r o x i m a t i o n to A changes before it settles down. Definition: A is n-r.e, iff there is a recursive a p p r o x i m a t i o n to A A0(x) such that for all = 0 A s x, and I{s : As(X) # As+l(X)} I ~ n. Note that this d e f i n i t i o n can be extended in an obvious w a y to apply to functions, predicate. too. The only 0-r.e. The n-r.e, [7], but note well that is quite different from ours.) set is ~ and the l-r.e, sets are the usual r.e. sets are those that arise after ting the Boolean algebra of r°e. A [8] calls an n-trial Similar ideas are p r e s e n t e d in Gold his d e f i n i t i o n of 2-r.e. sets. (This is what Putnam sets. Rogers n [9], p. 317 shows that is in the Boolean algebra g e n e r a t e d by the r.e. A < 0' (!btt -btt Putnam [8]. If G ~ 2~ means b o u n d e d - t r u t h - t a b l e denote by ~T the set steps in genera- sets iff reducible). {deg(A) : A See also E ~}. we always m e a n T u r i n g degree unless o t h e r w i s e specified. class of partial functions from T h e o r e m I: {A : A ~T any r.e° m o r e than any r.e. freedom we need. to is n - r . e . } T ~ {A : A To prove this we construct, w h i c h is ~ set. ~ define If ~ is a ~T = (~ N 2~) T. is n+l-r.e.} T. for e.g., n = I, a 2-r.e. We use the fact that set to execute By degree A the diagonalization. This proof is given in E p s t e i n set A may change once That's all the [4], A p p e n d i x 2, and 35 is due to R. W. Robinson. To the best of our knowledge, the first to prove this, Recalling whether X, X that in [2]. X, X n+l-r.e, r.e. implies that implies we'll m o d i f y the question A that a little. is w e a k l y n-r.e, X X is recursive we ask is n-r.e. To answer this Say that if A = lim s As(X) l{s : As(X) That is A0(x) = 0 Cooper was is no longer required. and # As+l(X)}l A picture will help: s 0 n-r.e. 0 i 2 3 4 x Diagram I I 2 0 0 0 0 0 S 3 4 0 I weakly n-r.e. 0 0 ~ n. i x 0 0 1 2 0 0 i 1 2 i 3 0 4 1 0 ~n changes allowed allowed Note that the weakly n-r.e, 4 6 i 1 i Pn changes all zero 3 R0, a recursive sets are closed under complementation: set just reverse the 0' and l's. The w e a k l y T h e o r e m 2: 0-r.e. A, A Proof: sets are the recursive are n+l-r.e, =] Easy. _~ Suppose approximations. iff they are weakly n-r.e. A = lim s As, At(x) all Then we'll only allow # At+l(X) x. Formally: A = lim s B s To obtain a w e a k l y n-r.e, we will go to the first stage at which C0(x) = As(X). # Bt+l(X) sets. approximation As(X) Cs(X) are n+l-r.e. # Bs(X) # Cs+l(X) C s to A, and set if we see since we know that in the end A(x) # B(x) Then there will be at most n changes. let s o = ~is (As(X) # Bs(X)) , and, Sm+ I = ps Then define (s > sm Cs(X) and = As0(X) for A s (x) # As(X) m all s i so • m < n+l, and As(X) # Bs(X)). 36 As(X) Cs(X) if sm ! s < Sm+ 1 = We leave to you that ASm+l(X) if Sm+ 1 ! s lim s Cs(X) = A(x). Clearly its n-r.e. An e s p e c i a l l y important fact about the r.e. degrees is that they are dense (see Sacks T h e o r e m 3: Given recursive r.e. Proof: procedure. such that out.) Let Cooper A What can we say about the n-r.e, n+l-r.e, and not recursive, If A there is some non- n = I; the rest follows by an inductive is r.e. w e ' r e done. F-E = A. f degrees? C iT A. We show this for (F So there are r.e. is the numbers put into be a I-I e n u m e r a t i o n of yourself that Corollary: [I0]). C F, is the required set. No n-r.e° and A, E sets E,F the ones taken C = f-l(E). Convince U degree is minimal. (unpublished) was the first to prove the Corollary, but by quite different means. By r e l a t i v i z i n g the proof of T h e o r e m 3 we can get that if 0 < d < a and ~ and b o t h are n+l-r.e, is r.e. in d. w h e t h e r the n-r.e, then there is some d < c < This does not solve for us the q u e s t i o n of degrees are dense: that's still open. us ask w h e r e in our c l a s s i f i c a t i o n schema such a We c, c But it makes will lie. I can extend our h i e r a r c h y by raising the bound on the n u m b e r of changes allowed. Definition: Given any A is f f-r.e, w h i c h is total iff there is some r e c u r s i v e a p p r o x i m a t i o n A s to A I{ s : As(X) such that # As+l(X)}I S i m i l a r l y we may define what it means for f-r. e. °.. c o n t i n u e d ! f(x). hiT 0' to be $7 We say a degree a is f-r.e, Note that this extends What do we know? Indeed A is f-r.e, the last place Let's As Certainly for some changes T h e o r e m 4. {A : A f iT 0': Proof: {A : A is f-r.e for some we can spot recursively class. it) in 0' Abbreviate identity- degree. a given r.e. argument for some ~= D(< 0') then is undecidable. a given r.e. # Cs+l(y) identity-r.e, is the construction set, by allowing y ! x. Bs(X) Any permitting Hence the construction degree produces an id-r.e, of a set B # Bs+l(X) only if argument produces an of a minimal degree below minimal degree (see e.g., [3]). All the degrees used in the proof in Epstein T h ( ( D ( < 0'), ~ ~ is undecidable Hence they are id-r.e., arithmetic goes through as for It is open w h e t h e r (see Soare are constructed T h ( ~ D ( < 0'); Th(({A : A is n-r.e.} T Th(${A : A is n+l-r.e.} T <7) : A Theorem 5: A Let arguof <7) is undecidable We can also ask w h e t h e r is the same as <)) sets. We say A is ~-r.e. if is f-r.e. $ be a class of total recursive we can recursively in by permitting <7). is r.e.} T Let us look at classes of f-r.e, f ~ ~, [4] and [5] that and the translation of fragments [12] for a discussion). Th(~{A some f. is not dense. is i d - r . e . } T = A permitting which is <T C, ments. definition. its mind. is id-r.e.} T Th(~; Epstein A iT 0' is f-r.e. as id-r.e. If Cs(Y) A ~ a the w e a k l y n-r.e, every look at a simple f-r.e, function-r.e, if some $. Let g enumerate functions such that an index for every function be a function which dominates every func- 38 tion in not 5. Then there is an A w h i c h is g-r.e, but is ~-r.e. The idea of the proof is just a m o d i f i c a t i o n of the p r o o f that there is a 2-r.e. an f-r.e, set degree w h i c h is not l-r.e. B, the set A w h i c h we are c o n s t r u c t i n g is allowed, for s u f f i c i e n t l y large x, The proof is in E p s t e i n [4], A p p e n d i x 2. Corollary: Here we k n o w that given to make one m o r e change than B does. Any of the usual h i e r a r c h i e s of r e c u r s i v e functions a h i e r a r c h y on the recursive-r.e, But the recursive-r.e, T h e o r e m 6: There is an induces degrees. degrees aren't the w h o l e story. A iT 0' for any r e c u r s i v e w h i c h is not f~r.e. f. The essential step in the proof is to show that we can "enumerate" all the recursive-r.e, approximations. [4], A p p e n d i x 2 and is due to Posner The proof appears in Epstein (private c o m m u n i c a t i o n 1973). How can we extend this hierarchy? Ershov [6]. 2 We take another tack as in Instead of b o u n d i n g the n u m b e r of changes by functions w h e n we pass from the n-r.e, C o n s i d e r that A case, we'll bound them by ordinals. being w e a k l y n-r.e, given by a collection of n can be v i e w e d as partial recursive functions, A being ~0 . . . . ' ~n" Here A(x) = ~k(X) Of course say that functions G where A k is the largest need not be 0-i valued. is w e a k l y n - a p p r o x i m a b l e ~0 . . . . ' ~n k = max t(~t(x)+). such that such that For any total A n d surely t ~t(x)+. G we could if there are partial recursive G(x) = ~k(X) How could we extend this to there needn't be a largest smallest one! t for w h i c h ~t(x)+. where w? In that case But there is a 39 {G : G sive is w e a k l y such After that G(x) all we m a y Definition: n-approximable} = ~k(X), collect ~ - - ÷ G, k = ~t(~t(x))+}. these ~ = {G : ~ 0 . . . . . ~i's partial as ~n partial Let's clean recurthis up. @((i,x)). recursive if for all x, (n) G(x) Now sive; it's then = ~(~k,x>) easy ~ ~ where to d e f i n e G if k = ~t < n ( ~ ( f t , x ~ ) ~ ) . an w-r.e, G(x) set. = ~(~k,x)) Let such @ be p a r t i a l that recur- k = ~t(~(~t,x))+). (~) From (~tt this means Let us definition truth-table it f o l l o w s that A is w-r.e, iff A < -tt 0' reducible). define V n = {A : some p a r t i a l recursive ~, ~ ~ A} (n) V = {A : some partial recursive ~, ~--+ A]. (~) Theorem 7: I. A E Vn+ 1 2. A E V Proof: x we our guess guess 2. idea iff We o u t l i n e d at n-I A(x) A where is simple: s i n c e we k n o w I. A A is w e a k l y is f-r.e, above. at m o s t n Note times, n-r.e. for some recursive that A ~ Vn iff which is the same f. for every as c h a n g i n g times. Suppose A = lim s A s ( x ) The iff that w e is f-r.e, the we changes reverse can for some label for recursive each x That are b o u n d e d the o r d e r of the our guess first f. labels by is by on the f(x). f(x). changes Formally, define undefined ¢(<r,x)) = if A0(x) if r = f(x) Ay(X) if r < f(x) was chosen to be y = ~t > z Now let us s u p p o s e that w e r > f(x) are g i v e n Az(X) (At(x) ~ ~ all computations truncated at stage s. @(<r+l,x)) and ~ Az(X)). A. (~) with and Define Let ~s D be 40 0 if s = 0 ~(~r,x>) A s if 0 Then l i m s As(X) r, as for all Corollary: Proof: Theorem class of sets There for a. 0 , }T ~ is no r e a s o n recursive. By As(X) makes for any w e n o t e that f q, must 0'} T " the o b s e r v a t i o n algebra generated V , yield that the f o r m e r by the r.e. to stop this h i e r a r c h y w i t h (see R o g e r s a-S [9], C h a p t e r w e have a r e c u r s i v e z = (Y)S II). ~. and ordering and s u c c e s s o r z a be any by a-S, systems distinguish we w i l l m e a n that Let Denote related univalent and the p r e d e c e s s o r sets, the c o r o l l a r y . w h e r e w e can r e c u r s i v e l y a n d limit o r d i n a l s , that ~s(q,x)+ r = f(x) {A : A J t t the list of r e q u r s i v e l y Thus g i v e n of c h a n g e s Zt ~((t,x>)+. is the B o o l e a n set of n u m b e r s , system that 5 and 7, plus ordinal and t h e r e is s u c h an for the first time Calling class of sets is constructive a-U . . . . x, : A Jbtt {A s > 0 A n d the n u m b e r where r = ~t ~ s ( ~ s ( < t , x ) ) + ) . the l a t t e r r = ~t j s ~ s ( < t , x > ) + otherwise = A(x). can't be m o r e t h a n be total where (x) = e-T, of n o t a t i o n on a r e c u r s i v e 0, s u c c e s s o r s functions are denotes T in the where y is the S. Definition: Given a-S, define iff (s-S) ~(x) = ~(((y)s,X>) least o r d i n a l < a s u c h that ~( <(~)s,X >)+. We r e a d ~ - + ~ as " ~ is the a-limit of in n o t a t i o n S." ~ in n o t a t i o n S" (a-S) or "~ a - a p p r o x i m a t e s notation we will often We n e e d not ~ delete suppose that reference ~ When working with to " n o t a t i o n is total a fixed S." in this d e f i n i t i o n . 41 Lastly define Va_ S = {~ : some partial recursive As the next theorem will demonstrate, be notation-dependent. n ! 0 and V We are not. as previously Corollary (to Theorem 7): See Theorem More importantly, Theorem 8: some of these classes will leave to the reader, defined. Th(fV( Ta-S); j >) Vn_ S = Vn, we really do have all the functions f iT 0 ' t h e n for some notation then and if a j w. a-S below 0' now, f iT 0' given any from Vn, 4. If a-S, to B-U f ~T 0'. As S, f ~ Vw2_S" we may find recursively Proof: 3 Suppose that for some informal proof that that is undecidable f ~ Va_ S that we ~an pass however, Immediately we can conclude If Moreover, ~ -~ ~} (a-S) That is, for any system S, V _S = V Proof: 4, a-S, S B-U, and B J a such f ~ VB_ U. f ~ Va_ S. We will give an is fixed we will suppress refer- ence to it throughout. Diagram 2 0 i .... w, ~+i .... B 0 1 2 ......... 3 4 x ..a . ~ . . . . . ~ ' ~ , . . . ~ . . ~ guess at priority f(x) with B; priority of guesses is well-ordered. To calculate if there is a If not, then f(x) look for some C < B such that ~((B,x>) = f(x). B with ,(<~,x>)+, ~(( S,x>)+. Then ask a question recursive If there is such a ~ in 0'. go to it and ask the same question. Since a is w e l l - o r d e r e d we can be sent to a smaller ~ with 42 ~((~,x>)+ then in only a finite number of times. f(x). This procedure for finding The least one we reach is f(x) is clearly recursive 0'. Note that if x, f(x)~ f ~ V iff f is partial the same proof works. we simply never make a guess at and is partial or total then Now assume that f ~ f ~T 0'. Thus if O We w i l l obtain a system of n o t a t i o n S such V~2_ S Since f iT 0', X = {<x,s> : s = 0 well-ordering (x = y f iT 0' f(x). Then for each and R f(x) = lim s f(s,x). or on f(s,x) # f(s+l,x)}. X s ! t)o Let given by Consider the recursive fx,s>R~y,t) iff x ! Y or This has order type w. Diagram 3 e.g. 012 . 32 , s 1 • 6 ) 54 8 13 7. 12 II i09 X As in the proof of T h e o r e m XX, Chapter Ii, Rogers [9] we obtain a sys- tem of n o t a t i o n as follows: xSy S for iff 2 by p a d d i n g out x = (v,n> and if and v # u, RIX y = (u,m> vRu; and <v,u> ~ X otherwise if v = u, n < m. Now define I f(s,x) ~(<<x,s>,x>) if = otherwise s = 0 or f(s,x) # f(s-l,x) 43 Clearly ~ ~ f. w -S Lastly, O given any e-S, we can o b t a i n a n o t a t i o n the same m a n n e r as d e s c r i b e d above, for w h i c h Note that if we w e r e to define classes only that e-R V~_ R for ~+~2-U a+~ • 2 f iT 0', there is some [] w h e r e we require R such that f ~ V _ R. G i v e n a n o t a t i o n ~-S. The classes For ~ = B+I vBT S for form a hierarchy. B ~ ~, That is, this is a m o d i f i c a t i o n of the proof of T h e o r e m 1 that there is a 2-r.e. set w h i c h allowed for each v a r i a b l e ~T x any l-r.e, set. A f t e r all, w e ' r e one more change than before. The changes in the proof are really only to accomodate the ordinal notation. consider ~ a limit ordinal. to construct an for each < ~. notation in is a recursive w e l l - o r d e r i n g on a recursive field, then given any T h e o r e m 9: f E V U x, A Essentially Now in this proof w e ' r e allowed w i t h an a p p r o x i m a t i o n w h o s e numbers of changes, dominates the n u m b e r of changes to any given Y ~ V~, for Thus the proof is, up to m o d i f i c a t i o n to accormmodate ordinal (which is not simple), the proof of T h e o r e m 5. Note that for m a n y - o n e degrees the h i e r a r c h y is better behaved: if A ~ m B ~ Va_ S then A ~ V _ S. This fails for T u r i n g degrees: by the C o r o l l a r y to T h e o r e m 3 any A iT 0' w h i c h has m i n i m a l degree provides a 4 c o u n t e r e x a m p l e for V I. T h e o r e m I0: Let If S A be a n o t a t i o n for ~ V .n_ S Conversely, W A is A x = y ~×n then is Vn-r.e. is the n o t a t i o n for o r d e r i n g on if if we have w-n. ~-n (namely r < t). Vn-r.e. then A ~ V .n_ W where given by the canonical (x,r) ( (y,t) iff x < y or 44 We will denote by the V .n class of Vn-r.e. functions. Proof: Let given by f ~ Vn We'll A ~ where which show ~ V .n_ S. at m o s t n informal proof and note can recursively (z) S = ~om+s, with f(x) leave that determine s deleting and .... 012 makes. recursively We present an that < k ordinals j ~-n (q)s = ~ - k + t or m = k. we and We proceed legibility. at We have first u-block we know f(0,x) = s. we in a n o t h e r and If w e stay We will for e a c h @ have time block. goes changes our changes in block y = ~-m+s is + (in a c o m p u t a t i o n As l o n g as w e its guess ~(~.x>)+ and ~ So s block change but now to c h a n g e ~ t this later have in t h i s .... in this m < n. can block, 012 0,.m+t which for s o m e that ....... most ~(<y,x>) ¥ = ~.m+s ~ at m o s t = in at the n copies f ~ once m for 0J ~ times to g u e s s notate such or subscripts function approximate correct. which t k < m that a total 4 012 are construct the i n f o r m a t i o n to the r e a d e r . q,z w Look are the d e t a i l s any via that we need until we given A of changes by showing and whether the p r o o f , Diagram at And we'll the number f E Vn times approximate ~ A. (~.n-S) dominates that First ~ We will at m o s t guess to a n o t h e r at th i n the m - - s times. ~ = w-k+t can change f(l,x) stay search). and k < m its m i n d at m o s t at m o s t n = s+t. f(x) ~-block. The final times, time So t 45 shifts is, say ~(~,x)+ f(r,x) = f(r-l,x) C = w.r+u, + u = f(x). Now suppose that our guess at and A(x) we have rl A = lim s As(X) is dominated by where f(0,x) . . . . , f(m,x) be the to f E Vn- A schematic p r e s e n t a t i o n will suffice Let the number of changes m to show that A E Vw. n. guesses we make at f(x), m < n-l. ~(x) the first kl-changes of As(X) where f(0,x) = kr ~I (x) k1 the next k2-changes of As(X) where f(l,x) = k 2. 2 (x) ~k2 A(x) = ~(x) the m th km-changes where f(m-l,x) = km n %k (x) m That is, we have for t j kr, ~(<(r,t>,x)) = ~kr_ tm-r (x). W Similarly we may prove Theorem II: If A E Vw.e_ S And if notates w×~ A is w.~ (namely (r) S < (t)s). then A V _s-r.e. is V _s-r.e. then A via the canonical ~x,r> < <y,t) iff E Vw.~_ W where ordering given by x < y W S or, if x = y, on 46 Corollary: For each i T 0'. such Proof: That that Given recursively e-S the from is, A s-r.e, given is not ~-S S V V sets S f-r.e, consider we [9], p. S" A is V _S,, r.e., The Theorem Using classes I0 w e canonical ~ ~ 6o6o. Vl-r. e . . . . V are out ... hence not V Vw. n ... V 2-r.e .... M. 0' obtained 9 we have Then V _ S r.e. Using e = ~-n any n. for any this. V6o.n-r. e . . . . vl V 2 "'" V 2 60 ,n 6O tT V A iT V 6o-r.e. 6O V sets as above. classes clarify 60 f-r.e. for t~ - - ~+I canonically. Vw-r. e . . . . 11 Vw S' the V _ S. By T h e o r e m canonical will some for from given classes A picture Vn-r. e . . . . 11 obtained ~ ~ 6o ii w e m a y n o w p i c k where f-r.e. for can g i v e Theorem = w.B V f ~ S' 206). A E Vw.(~+I)_S,,-V~.~_S,,, exhaust find for any an but not can a notation (see R o g e r s V +I_S,, r.e. do not ~t . 3 Lerman . V . has 6o recently communicated to us that he and L. H a y have proved: for all that He suggests proof shown r.e. the that of T h e o r e m Also R. n ! 1 interval the p r o o f there degree b, {b are two n+l-r.e, : c < b < d} is a not difficult degrees ~ has no n-r.e, modification < d such degrees. of the i. Shore that there and L. H a y is a 2 r.e. a < b < O' have degree communicated a < O' to us such that that they have there is no 47 I. Hay and Lerman have observed prove that if which is a > 0 m+l-r.e, is r.e. that by a permitting then for and not m-r.e., Vm+l 2. is f-r.e, to for any n ! i. from the fact that And that appears Arch. n-r.e, b A !tt 0' iff Math Logik 18(1976), 55-65. the same as T h e o r e m 5 and 6(2) of it fails for every for every r.e. [5]). a Vn there is some by Theorem 3 and the fact that m < a A as Theorem 2,3 in [6] (part II). Actually Epstein f by H. G. Carstens, This proof is essentially Ershov A itt 0' . but for some recursive "A~-Mengen" 4. A iT 0' A corollary is L. Hay has pointed out that Theorem 6 follows there is an 3. a one can there is some a > b > 0. Theorem 3 is that the same is true if argument of minimal degree (see 48 Bibliography [I] Addison, J., The method of alternating chains, in T h e o r y M o d e l s , North-Holland, Amsterdam, 1965 (p. 1-16). [2] Cooper, S. B., Doctoral Dissertation, University of Leicester, 1971. [3] Epstein, Richard L., M i n i m a l Full A p p r o x i m a t i o n 1975. of D e g r e e s of U n s o l v a b i l i t y and the Construction, Memiors of the A.M.S., no. 162, [4] Epstein, Richard L., D e g r e e s of U n s o l v a b i l i t y : S t r u c t u r e and T h e o r y , Lecture Notes in Mathematics no. 759, Springer-Verlag, New York. [5] Epstein, Richard L., I n i t i a l [6] Ershov, A. Hierarchy of Sets I, II, III, A l g e b r a and L o g i c , VOI. 7, no. i, no. 4 (1968) and vol. 9, no. 1 (1970). (English translation, Consultants Bureau, N.Y.). [7] Gold, Limiting recursion, J o u r n a l no. I, p. 28-48, 1965. [8] Putnam, H., Trial and error predicates and the solution to a problem of Mostowski, J o u r n a l of S y m b o l i c Logic, vol. 30, no. I, p. 49-57, 1965. [9] Rogers, Hartley, T h e o r y of R e c u r s i v e Computation, McGraw-Hill, New York. [I0] [II] segments of Degrees of Symbolic Functions < 0', to appear. Logic, and VOI. 30, Effective Sacks, Gerald, D e g r e e s o f U n s o l v a b i l i t y , Annals of Math. Studies, no. 55, Princeton, New Jersey, revised edition, 1965. Shoenfield, J. R., On the degrees of unsolvability, vol. 69, p. 644-653, 1959. Annals of Mathematics, [12] Soare, Robert, Recursively enumerable sets and degrees, Bull. VOI. 84, no. 6 (1978), p. 1149. A.M.S., Iowa State University Ames, lowa 50011 University of California Berkeley, California 94720 Iowa State University Ames, Iowa 50011