diff --git a/src/contexts.jl b/src/contexts.jl
deleted file mode 100644
index 706e579..0000000
--- a/src/contexts.jl
+++ /dev/null
@@ -1,64 +0,0 @@
-abstract type AbstractContext
-
-
-"""
-    struct JointContext <: AbstractContext end
-
-The `JointContext` is used to evaluate a probabilistic program as a joint distribution over data
-and parameters.
-"""
-struct JointContext <: AbstractContext end
-
-
-"""
-    struct PriorContext{Tvars} <: AbstractContext
-        vars::Tvars
-    end
-
-The `PriorContext` is used to evaluate the prior distribution of a conditioned probabilistic model 
-(i.e., the unconditional distribution of the parameters `vars`).
-"""
-struct PriorContext{Tvars} <: AbstractContext
-    vars::Tvars
-end
-PriorContext() = PriorContext(nothing)
-
-
-"""
-    struct LikelihoodContext{Tvars} <: AbstractContext
-        vars::Tvars
-    end
-
-The `LikelihoodContext` is used to evaluate the likelihood distribution of a conditioned
-probabilistic model (i.e., the conditional distribution of the data given the the parameters).
-
-If `vars` is `nothing`, the parameter values inside the `VarInfo` will be used by
-default.
-"""
-struct LikelihoodContext{Tvars} <: AbstractContext
-    vars::Tvars
-end
-LikelihoodContext() = LikelihoodContext(nothing)
-
-
-"""
-    struct MiniBatchContext{Tctx, T} <: AbstractContext
-        ctx::Tctx
-        scale::T
-    end
-
-The `MiniBatchContext` wraps another context to modify the joint log-density as `log(prior) + scale
-* log(likelihood of a batch)` when running the model, where `scale` is typically the
-number of data points divided by the batch size.  (Since this affects only the likelihood, it may be
-applied to joint and likelihood contexts).
-
-This is useful in batch-based stochastic gradient descent algorithms which optimize `log(prior) +
-log(likelihood of all the data points)` in the expectation.
-"""
-struct MiniBatchContext{Tctx, T} <: AbstractContext
-    ctx::Tctx
-    scale::T
-end
-function MiniBatchContext(ctx = JointContext(); batch_size, npoints)
-    return MiniBatchContext(ctx, npoints / batch_size)
-end