-*- mode:text; coding:utf-8 -*-

                                    March 23, 1959

Artificial intelligence Project---RLE and MIT Computation Center

                         Memo 8


                       BY MACHINE

                     by J. McCarthy

                   An Error in Memo 8

     The definition of eval given on page 15 has two errors,
one of which is typographical and the other conceptual.  The
typographical error is in the definition of evcon where
"1⟶" and "T⟶" should be interchanged.
     The second error is in evlam.  The program as it stands
will not work if a quoted expressoin contains a symbol which
also acts as a variable bound by the lambda.  This can be
corrected by using insteaad of subst in evlam a function subsq
defined by

                                    March 4, 1959

Artificial intelligence Project---RLE and MIT Computation Center

                         Memo 8


                       BY MACHINE

                     by J. McCarthy

     The attached paper is a description of the LISP system
starting with the machine-independent system of recursive
functions of symbolic expressions.  This seems to be a better
point of view for looking at the system than the original
programming approach.  After revision, the paper will be sub-
mitted for publication in a logic or computing journal.
     This memorandum contains only the machine independent
parts of the system.  The representation of S-expressions in
the computer and the system for representing S-functions by
computer subroutines will be added.


       by J. McCarthy, MIT Computation Center

1.    Introduction
      A programming system called LISP (for LISt Processor)
has been developed for the IBM 704 computer by the Artificial
Intelligence Group at MIT.  The system was designed to facili-
tate experiments with a proposed system called the Advice Taker
whereby a machine could be instructed in declarative as well
as imperative sentences and could exhibit "common sense" in
carrying out its instructions.  The original proposal for the
Advice Taker is contained in reference 1.  The main require-
ment was a programming system for manipulating    expressions
representing formalized declarative and imperative sentences
so that the ADvice Taker system could make deductions.
     The development of the LISP system went through several
stages of simplification in the course of its development and
was eventually seen to be based on a scheme for representing
the partial recursive functions of a certain class of symbolic
expressions.  This representation is independent of the IBM 704
or any other electronic computer and it now seems expedient
to expound the system starting with the class of expressions
called S-expressions and the functions called S-functions.
     In this paper, we first describe the class of S-expressions
and S-functions.  Then we describe the representation of
S-functions by S-expressions which enables us to prove that
all computable partial functions have been obtained, to obtain
a universal S-function, and to exhibit a set of questions
about S-expressions which cannot be decided by an S-function.
We describe the representation of the system in the IBM 704,
including the representation of S-expressions by list structures
similar to those used by Newell, Simon, and Shaw (see refer-
ence 2), and the representation of S-functions by subroutines.
Finally, we give some applications to symbolic calculations
including analytic differentiation, proof checking, and
compiling including a description of the present status of
the LISP compiler itself which is being written in the system.
     Although we have not carried out the development of


recursive function theory in terms of S-functions and their
representation by S-expressions beyond the simplest theorems,
it seems that formulation of this theory in terms of S-func-
tions has important advantages.  Devices such as Gödel number-
ing are unnecessary and so is the construction of particular
Turing machines.  (These constructions are all artificial in
terms of what is intended to be accomplished by them).  The
advantage stems from the fact that functions of symbolic
expressoins are easily and briefly described as S-expressions
and the representation of S-functions by S-expressions is
trivial.  Moreover, in a large class of cases the S-expression
representations of S-functions translate directly into effi-
cient machine programs for the computation of the functions.
Although, the functions described in the manner of this paper
include all computable functions of S-expressions, describe
many important processes in a very convenient way, and compile
into fast running programs for carrying out the processes;
there are other kinds of processes whose description by S-func-
tions is inconvenient and for which the S-functions once found
do not naturally compile into efficient programs.  For this
reason, the LISP system includes the possibility of combining
S-functions into Fortran or IAL-like programs.  Even this
will not provide the flexibility of description of processes
hoped for from the Advice Taker system which is beyond the
scope of this paper.

2.   Recursive Functions of Symbolic Expressions
     In this section we define the S-expressions and the
S-functions.  (Actually they are partial functions.  The
distinction between a partial function and a function that
the former need not be defined for all arguments because, for
example, the computation process defining it may not terminate.)

     2.1. S-expressions
     The expression with which we shall deal are formed
using the special characters "," and "("  and  ")" and an
infinite set of distinguishable atomic symbols p₁,p₂,p₃,....


     The S-expressions are formed according to the following
recursive rules.
          1.   The atomic symbols p₁ p₂ etc are S-expressions.
          2.   A null expression ⋀ is also admitted.
          3.   If e is an S-expression so is  (e).
          4.   If e₁ and (e₂) are S-expressions so is (e₁,e₂).
     In what follows we shall use sequences of capital Latin
letters as atomic symbols.  Since we never juxatpose them with-
out intervening commas they cannot cause confusion by running
together.  Some examples of S-expressions are;
     2.2  Elementary functions and predicates
     The functions we shall need are built up from certain
elementary ones according to certain recursive rules.
     There are three elementary predicates:
     1.   null[e]
          null[e] is true if and only if S-expression e is the
null expression ⋀.  (We shall use square brackets and semi-colons
for writing functions of S-expressions since parentheses and
commas have been pre-empted.  When writing about functions in
general we may continue to use parentheses and commas.)
     2.   atom[e]
          atom[e] is true if and only if the S-expression is
an atomic symbol.
     3.   p₁=p₂
          p₁=p₂ is defined only when p₁ and p₂ are both atomic
symbols in which case it is true if and only if they are the
same symbol.  This predicate expresses the distinguishability
of the symbols.
     There are three basic functions of S-expressions whose
values are S-expressions.
     4.   first[e]
          first[e] is defined for S-expressions which are
neither null nor atomic.  If e has the form (e₁,e₂) where e₁
is an expression, then first[e]=e₁.  If e has the form (e₁)
wehre e₁ is an S-expression again we have first[e]=e₁.


          Some examples are:
          first[A] is undefined
     5.   rest[e]
          rest[e] is also defined for S-expressions which are
neither null nor atomic.  If e has the form (e₁,e₂) where e₁
is an S-expression, then rest[e]=(e₂).  If e has the form (e₁)
wehre e₁ is an S-expression we have rest[e]=⋀.
          Some examples are:
     6.   combine[e₁;e₂]
          combine[e₁;e₂] is defined when e₂ is not atomic.
                When e₂  has the form (e₃), then combine[e₁;e₂]=(e₁,e₃)
     When e₂ is ⋀ we have combine[e₁;e₂]=(e₁).
          Some examples are:
     The functions first, rest and combine are related by the
whenever all the quantities involved are defined.

     2.3  Functional Expressions and Functions formed from the
     elementary functions by composition.
          Additional functions may be obtained by composing the
elementary functions of the preceding section.  These functions
are decribed by expressions in the meta-language which should
not be confused with the S-expressions being manipulated.  For
example, the expression first[rest[e]] debute the second sub-
expression of the S-expression e, e.g. first[rest[(A,B,C)]]=B.
In general compositions of first and rest give sub-expressions of
an S-expression in a given position within the expression and


compositions of combine form S-expressions from their sub-
expressions.  For example, combine[x;combine[y;combine[z,⋀]]]
forms as sequence of three ters from the terms, e.g. combine
     In order to be able to name compositions of functions and
not merely functional expressions (forms) we use the Church
λ-notation.  If ℰ is a functional expression and x₁,...,xn are
variables which may occur in ℰ, then λ[[x₁,...,xn],] denotes
the function of n variables that maps x₁,...,xn into ℰ.  For
example, λ[[x],first[rest[x]]] is the function which selects the
second element of a list and we have λ[[x],first[rest[x]]][(A,
[B,C],A)]=[B,C].  λ[[x];[A,B]] is the constant function that
maps every S-expression into [A,B].
     The variables occuring in the list of a λ-expression are
bound and replacing such a variable throughout a λ-expression
by a new variable does not change the function represented.
Thus λ[[x,y], combine[x,combine[y,⋀]]] is the same function as
λ[[u,v], combine[u,combine[v,⋀]]] but different from λ[[y,x],
     If some of the variables in a functional expressoin or
form are bound by λ's and others are not, we get a function
dependent on parameters or from another point of view a form
whose value is a function when values have been assigned to
the variables.
     2.4  Conditional Expressions
     Let p₁,p₂,...,pk be expressions representing propositions
and let e₁,...,ek be arbitrary expressions.   The expression
[p₁⟶e₁,...pk⟶ek] is called a conditional expression and its
value is determined from the values assigned to the variables
occuring in it as follows:  If the value of p₁ is not defined
neither is that of the conditional expression.  If p₁ is defined
and true the value of the conditional expression is that of e₁
if the latter is defined and otherwise is undefined.  If p₁ is
defined and false, then the value of [p₁⟶e₁,...,pk⟶ek] is that
of [p₂⟶e₂,...pk⟶ek].  Finally if pk is false the value of
[pk⟶ek] is undefined.
     An example of a conditional expression is [null[x]⟶⋀;
atom[x]⟶⋀;1⟶first[x]].  The "1" occuring in the above
expression is the propositional constant "truth".  We also use


"0" for the propositional constant "falsehood".  When used as then
last proposition in a conditional expression "1" may be read
"in all remaining cases".  The expression given is a sort of
extension of the expressoin first[x] which is defined for all
S-expressions.  We could define a corresponding function by
first_a=λ[[x]; [null[x]⟶⋀;atom[x]⟶⋀;1⟶first[x]]].
     It is very important to note that for a conditional
expressoin to be defined it is not necessary for all of its
sub-expressions to be defined.  If p₁ is defined and true and
e₁ is defined, the conditional expression [p₁⟶e₁,...,pk⟶ek]
is defined even if none of the other p's or e's is defined.
If p₁ is defined and false, p₂ is defined and true and e₂ is
defined, the expression is defined even if e₁ and all the other
p's and e's are undefined.
     The propositional connectives ∧ and ∨ and ∼ may be defined
in terms of conditional expressions.  We have p₁∧p₂=[p₁⟶[p₂
⟶1,1⟶0],1⟶0] and p₁∨p₂=[p₁⟶1,p₂⟶1,1⟶0] and ∼p=[p⟶0,1⟶1]
There is a slight difference between the connecetives defined
this way and the ordinary connectives.  Suppose that p₁ is
defined and true but p₂ is undefined.  Then p₁∨p₂ is defined
and true but p₂∨p₁ is undefined.
     2.5  Recursive Function Definitions
     The functions which can be obtained from the elementary
functions and predicates by composition and conditional expres-
sions form a limited class. As we have decribed them they are
not defined for all S-expressions but if we modified the
definitions of the elementary functions so that the undefined
cases are defined in some trivial way, as in the example of the
pervious section, the would be always defined.
     Additional functions may be defined by writing definitions
of the form,
     f=λ[[x₁,...,xn],ℰ] where the expression ℰ may contain the
symbol f itself.  A function f defined in this way is to be
computed for a given argument is to be computed by substitution
of the argument into the expression and attempting to evaluate
the resulting expression.  When a conditional expression is
encountered we evaluate p's until we find a true p and then
evaluate the corresponding e.  No attempt is made


to evaluate later p's or any e except the one corresponding
to the first e.  It may happen that in evaluating for given
values of the variables it is unnecessary to evaluate any
expression involving the defined function f.  In this case,
the evaluation may be completed and the function defined for
this argument.  If expressions involving f do have to be
evaluated we substitute the arguments of f and again proceed
to evaluate.  The process may or may not terminate.  For
those arguments for which the process does terminate the
function is defined.
     We shall illustrate this concept by several examples:
     1.  Our first example is a function which gives the
first symbol of an expression:
We defined
Let us trace the computation of ff[(((A),B),C)].  We have
*    Note that it does not matter that first A occuring in the
next to last step is undefined.
     2.   The second example is a function which gives the result
of substituting the expression x for the symbol y in the expres-
sion s.  We define
     We shall illustrate teh application of this definition by
computing subst[(A,B);X;((X,A),C)].  In order to make the tracing
shorter we shall give the situation at each recursion and leave
it to the reader to substitute the definition of each subst
expression and to check the determination of which case of the


conditional is applicable.  We have
     2.6  Functions with Functions as arguments
     If we allow variables representing functions to occur in
expressions and create functions by incorporating these variables
as arguments of λ's we can define certain functions more concisely
than without this facility.  However, as we shall show later no
additional S-function become definable.
     As an example of this facility we define a function maplist [x,f]
     where x is an S-expression and f is a function from S-expres-
sions to S-expressions.  We have
     The usefulness of maplist is illustrated byb formulas for
the partial derivative with respect to x of expressions involving
sums and products of x and other variables.  The S-expressions
we shall differentiate are formed as follow:
1.   An atomic symbol is an allowed expression.
2.   If e₁;e₂;...;en are allowed expressions so are (PLUS,e₁,
...,en) and (TIMES,e₁,...,en) and represent the sum and product
respectively of e₁;...;en
     This is essentially the Polish notation for functions except
that the inclusion of parentheses and commas allows functions of
variable numbers of arguments.  An example of an allowed
expression is
the conventional algebraic notation for which is X(X+A)Y
     Our differentiation formula is


     The derivative of the above expression computed by this
formula is
     2.7  Labelled Expressions
     The λ-notation used for naming functions is inadequate
for naming recursive functions.  For example, if the function
named as the second argument of a maplist is to be allowed to
be recursive an additional notation is required.
     We define label[s;e] where  s  is a symbol and  e  is an
expression to be the same as the exression  e  except that
if  s  occurs as a sub-expression of  e  it is understood to
refer to the exression  e.  The symbol  s  is bound in label
[s;e]  and has no significance outside this expression.  Label
acts as a quantifier with respect to its first argument but a
quantifier of a different sort from λ.  As an example
is a name suitable for inclusion in a maplist of the substitu-
tion function mentioned earlier.
     2.8  Computable Functions
     In this section we shall show that all functions compu-
table by Turing machine are expressable as S-functions.  If,
as we contend, S-functions are a more suitable device for
developing a theory of computability than Turing machines,
the proof in this section is out of place and should be re-
placed by a plausibility argument similar to what is called
"Turing's thesis" to the effect that S-functions satisfy our
intuitive notion of effectively computable functions.  The
reader unfamiliar with Turing machines should skip this section.
     Nevertheless, Turing machines are well entrenched at present
so we shall content ourselves with showing that any function
computable by turing machine is an S-function.  This is done
as follows:
     1.   We give a way of describing the instantaneous con-
figurations of a Turing machine calculation by an S-expression.
This S-expression must describe the turing machine, its


internal state, the tape, and the square on the tape being
     2.   We give an S-function succ whose arguments is an
instantaneous configuration and whose value is the immediately
succeeding configuration if there is one and otherwise is 0.
     3.   We construct from succ another S-function, turing,
whose arguments are a Turing machine, with a canonical initial
state and an initial tape in a standard position and whose
value is defined exactly when the corresponding Turing machine
calculation terminates and in that case is the final tape.
     We shall consider Turing machines as given by sets of
quintuples.  Each quintuple consists of a state,
a symbol read, a symbol to be printed, a direction of motion
and a new state.  The states are represented by a finite set
of symbols, the symbols which may occur by another finite set
of symbols (it doesn't matter whether these sets overlap) and
the two directions by the symbols "L" and "R".  A quintuple is
then represented by an S-expression (st,sy,nsy,dir,nst).
The Turing machine is represented by an S-expression. (ist,
blank,quin1,...quink) were ist represents the canonical
initial state, blank is the symbol used for a blank square
(squares beyond the region explicitly represented in the
S-expression for a tape are assumed to be blank and are read
that way when reached).  As an example, we give the representa-
tion of a Turing machine which moves to the right on a tape
and computes the parity of the number of 1's on the tape ignoring
0's and stopping when it comes to a blank square:
The machine is assumed to stop if there is no quintuple with a
given symbol state pair so that the above machine stops as soon
as it enters state 2.
     A Turing machine tape is represented by an S-expression as
follows:  The symbols on the squares to the right of the scanned
square are given in a list v, the symbols to the left of the
scanned square in a list u and the scanned symbol as a quantity w.
These care combined in a list (w,u,v).  This the tape ...bb1101⓪10b...
is represented by the expression


     We adjoin the state to this triplet to make a quadruplet
(s,w,u,v) which describes the instantaneous configuration of a
     The function succ[m;c] whose arguments are a Turing machine
m  and  a configuration  c  has as value the immediately suc-
ceeding configuration of  c  provided the state-symbol pair is
listed among the quintuplets of  m  and otherwise has value zero.
     succ is defined with the aid of auxiliary functions.  The
first of these  find[st;sy;qs] given the triplet (nsy;dir;nst)
which consists of the last 3 terms of the quintuplet of  m  which
contains  (st,sy)  as its first two elements.  The recursive
definition is simplified by defining find [st;sy;qs] where
qs=rest[rest[m]] since qs then represents the list of quintuplets
of  m.  We have find[st;sy;qs]=[null[qs]⟶0;first[first[qs]]
     The new auxiliary function is move[m;nsy;tape;dir] which
gives a new tape triplet obtained by writing nsy  on the scanned
square of tape, and moving in the direction dir.
     The reader should not be alarmed at the monstrous size of
the last formula.  It rises mainly from the compositions of first
and rest required to select the proper elements of the structure
representing the tape. Later we shall descrirbe ways of writing
such expressions more concisely.
     We now have


     Finally we define
     We reiterate that these definitions can be greatly shortened
by some devices that will be discussed in the sections on the
machines  computation of S-functions.


3.   Lisp Self-applied
     The S-functions have been described by a class of expres-
sions which has been informally introduced.  Let us call these
expressions F-expressions.  If we provide a way of translating
F-expressions into S-expressions, we can use S-functions to
represent certain functions and predicates of S-expressions.
     First we shall describe this translation.
     3.1  Representation of S-functions as S-expressions.
     The representation is determined by the following rules.
     1.  Constant S-expressions can occur as parts of the
F-expressions representing S-functions.  An S-expression ℰ is
represented by the S-expression.  (QUOTE,ℰ)
     2.  Variables and function names which were represented
by strings of lower case letters are represented by the cor-
responding strings of the corresponding upper case letters.
Thus we have FIRST, REST and COMBINE, and we shall use X,Y
etc. for variables.
     3.  A form is represented by an S-expression whose first
term is the name of the main function and whose remaining terms
are the argumetns of the function.  Thus combin[first[x];
rest[x]] is represented by (COMBINE,(FIRST,X),(REST,X))
     4.  The null S-expression ⋀ is named NIL.
     5.  The truth values  1  and 0  are denoted by T and F.
         The conditional expressoin
is repersented by
     6.  λ[[x;..;s];ℰ] is represented by (LAMBDA,(X,...,S),ℰ)
     7.  label[α;ℰ] is represented by (LABEL,α,ℰ)
     8.  x=y is represented by (EQ,X,Y)
     With these conventions the substitution function mentioned
earlier whose F-expression is
is represented by the S-expression.


     This notation is rather formidable for a human to read,
and when we come to the computer form of the system we will
see how it can be made easier by adding some features to the
read and print routines without changing the internal compu-
tation processes.
     3.2.  A Function of S-expressions which is not an S-function.
     It was mentioned in section 2.5 that an S-function is not
defined for values of its arguments for which the process of
evaluation does not terminate.  It is easy to give examples
of S-functions which are defined for all arguments, or examples
which are defined for no arguments, or examples which are
defined for some arguments.  It would be nicie to be able to
determine whether a given S-function is defined for given
arguments.  Consider, then, the function def[f;s] whose value
is 1 if the S-function whose corresponding S-expression is  f
is defined for the list of arguments  s  and is zero otherwise.
     We assert that def[f,s] is not an S-function.  (If we
assume Turing machine theory this is an obvious consequence
of the results of section 2.8, but in support of the contentions
that S-functions are a good vehicule for expounding the theory
of recursive functions we give a separate proof).

Theorem:  def[f;s] is not an S-function.
Proof:    Suppose the contrary.  Consider the function
     If def were an S-function g would also be an S-function.
For any S-function  u  with S-expression  u' g[u'] is 1 if u[u']
undefined and is undefined otherwise.
Consider now g[g'] where g' is an S-expression for g.  Assume
first that g[g'] were defined.  This is precisely the condi-
tion that g' be the kind of S-expression for which  g  is
undefined.  Contrawise, were g[g'] undefined  g'  would be the
kind of S-expression for which  g  is defined.
     Thus our assumption that def[f;s] is an S-function leads
to a contradiction.
     The proof is the same as the corresponding proof in
Turing machine theory.  The simplicity of the rules by which
S-functions are represented as S-expressions makes the develop-
ment from scratch simplier, however.


     3.3  The Universal S-Function, Apply
     There is an S-function apply such that if  f  is an
S-expression for an S-function φ and args is a list of the
form (arg1,...,arg n) where arg1,---,arg n are arbitrary
S-expressions then apply[f,args] and φ[arg1;...;argn]
are defined for the same values of arg1,...arg n and are
equal when defined.
     apply is defined by
     eval is defined by
where: evcon[c]=[eval[first[first[c]]]=1⟶eval[first[rest[first[c]]]];
    The proof of the above assertion is by induction on the
subexpressions of  e.  The process described by the above
functions is exactly the process used in the hand-worked
examples of section 2.5.


4.   Variants of Lisp
     There are a number of ways of defining functions of
symbolic expressions which are quite similar to the system
we have adopted.  Each of them involves three basic functions,
conditional expressions and recursive function definitions,
but the class of expressions corresponding to S-expressions
differs and sod o the precise definitions of the functions.
We shall describe two fo these variants.
     4.1  Linear Lisp
          The L-expressions are defined as follows:
     1.   A finite list of characters is admited.
     2.   Any string of admited characters in an L-expres-
sion.  This includes the null string denoted by ⋀
     There are three functions of strings
     1.   first[x] is the first character of the string x.
          first[⋀] is undefined.
     For example, first [ABC]=A.
     2.   rest[x] is the string of characters remaining when
the first character of the string is deleted.
          rest[⋀] is undefined.
          For example, rest[ABC]=BC
     3.   combine[x;y] is the string formed by prefixing the
character  x  to the string  y.
          For example, combine[A;BC]=ABC
     There are three predicates on strings
     1.   char[x],  x is a single character
     2.   null[x],  x is the null string
     3.   x=y,      defined for  x  and  y characters.
     The advantage of linear Lisp is that no cahracters are
given special roles as are parentheses and comma in Lisp.
This permits computations with any notation which can be
written linearly.  The disadvantage of linear Lisp is that
the extraction of sub-expressions is a fairly involed rather
than an elementary operation.  It is not hard to write in
linear lisp functions corresponding to the basic functions of
Lisp so that mathematically linear Lisp includes Lisp.  This
turns otu to be the most convenient way of programming more
complicated manipulations.  However, it tunrs out that if the
functions are to be represented by computer routines Lisp is
essentially faster.


     4.2  Binary Lisp
     The unsymmetrical status of  first  and  rest  may be a
source of uneasiness.  If we admit only two element lists
then we can define
     We need only two predicates, equality for symbols and
atom.  The null list can be dispensed with.  This system is
easier until we try to represent functions by expressions which
is, after all, the principal application; moreover, in order
to apply the system to itself we need to be abel to write