diff --git a/Project.toml b/Project.toml
index 01a80cd..31ac1c3 100644
--- a/Project.toml
+++ b/Project.toml
@@ -1,7 +1,7 @@
 name = "GenParticleFilters"
 uuid = "56b76ac4-72ef-411e-b419-6d312ed86a6f"
 authors = ["Xuan <tanqazx@gmail.com>"]
-version = "0.1.4"
+version = "0.1.5"
 
 [deps]
 Distributions = "31c24e10-a181-5473-b8eb-7969acd0382f"
diff --git a/src/GenParticleFilters.jl b/src/GenParticleFilters.jl
index 706cbb7..953d456 100644
--- a/src/GenParticleFilters.jl
+++ b/src/GenParticleFilters.jl
@@ -3,8 +3,9 @@ module GenParticleFilters
 using Gen, Distributions
 using Gen: ParticleFilterState
 
-export ParticleFilterState
+export ParticleFilterState, ParticleFilterSubState, ParticleFilterView
 
+include("view.jl")
 include("utils.jl")
 include("initialize.jl")
 include("update.jl")
diff --git a/src/rejuvenate.jl b/src/rejuvenate.jl
index bf15b6b..020bd03 100644
--- a/src/rejuvenate.jl
+++ b/src/rejuvenate.jl
@@ -12,7 +12,7 @@ a tuple with a trace as the first return value. `method` specifies the
 rejuvenation method: `:move` for MCMC moves without a reweighting step,
 and `:reweight` for rejuvenation with a reweighting step.
 """
-function pf_rejuvenate!(state::ParticleFilterState, kern, kern_args::Tuple=(),
+function pf_rejuvenate!(state::ParticleFilterView, kern, kern_args::Tuple=(),
                         n_iters::Int=1; method::Symbol=:move)
     if method == :move
         return pf_move_accept!(state, kern, kern_args, n_iters)
@@ -34,7 +34,7 @@ a tuple `(trace, accept)`, where `trace` is the (potentially) new trace, and
 can be supplied with `kern_args`. The kernel is repeatedly applied to each trace
 for `n_iters`.
 """
-function pf_move_accept!(state::ParticleFilterState,
+function pf_move_accept!(state::ParticleFilterView,
                          kern, kern_args::Tuple=(), n_iters::Int=1)
     # Potentially rejuvenate each trace
     for (i, trace) in enumerate(state.traces)
@@ -44,10 +44,7 @@ function pf_move_accept!(state::ParticleFilterState,
         end
         state.new_traces[i] = trace
     end
-    # Swap references
-    tmp = state.traces
-    state.traces = state.new_traces
-    state.new_traces = tmp
+    update_refs!(state)
     return state
 end
 
@@ -66,7 +63,7 @@ accumulated accordingly.
 [1] R. A. G. Marques and G. Storvik, "Particle move-reweighting strategies for
 online inference," Preprint series. Statistical Research Report, 2013.
 """
-function pf_move_reweight!(state::ParticleFilterState,
+function pf_move_reweight!(state::ParticleFilterView,
                            kern, kern_args::Tuple=(), n_iters::Int=1)
     # Move and reweight each trace
     for (i, trace) in enumerate(state.traces)
@@ -79,10 +76,7 @@ function pf_move_reweight!(state::ParticleFilterState,
         state.new_traces[i] = trace
         state.log_weights[i] += weight
     end
-    # Swap references
-    tmp = state.traces
-    state.traces = state.new_traces
-    state.new_traces = tmp
+    update_refs!(state)
     return state
 end
 
diff --git a/src/resample.jl b/src/resample.jl
index 70ba79d..0243d06 100644
--- a/src/resample.jl
+++ b/src/resample.jl
@@ -60,10 +60,7 @@ function pf_multinomial_resample!(state::ParticleFilterState;
         ws = state.log_weights[state.parents] .- log_priorities[state.parents]
         state.log_weights = ws .+ (log(n_particles) - logsumexp(ws))
     end
-    # Swap references
-    tmp = state.traces
-    state.traces = state.new_traces
-    state.new_traces = tmp
+    update_refs!(state)
     return state
 end
 
@@ -115,10 +112,7 @@ function pf_residual_resample!(state::ParticleFilterState;
         ws = state.log_weights[state.parents] .- log_priorities[state.parents]
         state.log_weights = ws .+ (log(n_particles) - logsumexp(ws))
     end
-    # Swap references
-    tmp = state.traces
-    state.traces = state.new_traces
-    state.new_traces = tmp
+    update_refs!(state)
     return state
 end
 
@@ -170,9 +164,6 @@ function pf_stratified_resample!(state::ParticleFilterState;
         ws = state.log_weights[state.parents] .- log_priorities[state.parents]
         state.log_weights = ws .+ (log(n_particles) - logsumexp(ws))
     end
-    # Swap references
-    tmp = state.traces
-    state.traces = state.new_traces
-    state.new_traces = tmp
+    update_refs!(state)
     return state
 end
diff --git a/src/update.jl b/src/update.jl
index 4d57b23..ced806f 100644
--- a/src/update.jl
+++ b/src/update.jl
@@ -9,7 +9,7 @@ Perform a particle filter update, where the model arguments are adjusted and
 new observations are conditioned upon. New latent choices are sampled from
 the model's default proposal.
 """
-function pf_update!(state::ParticleFilterState, new_args::Tuple,
+function pf_update!(state::ParticleFilterView, new_args::Tuple,
                     argdiffs::Tuple, observations::ChoiceMap)
     n_particles = length(state.traces)
     for i=1:n_particles
@@ -20,10 +20,7 @@ function pf_update!(state::ParticleFilterState, new_args::Tuple,
         end
         state.log_weights[i] += increment
     end
-    # Swap references
-    tmp = state.traces
-    state.traces = state.new_traces
-    state.new_traces = tmp
+    update_refs!(state)
     return state
 end
 
@@ -51,7 +48,7 @@ that occur in `a` also occur in `b`, and the values at those addresses are
 equal. It is an error if no trace `t_new` satisfying the above conditions
 exists in the support of the model (with the new arguments).
 """
-function pf_update!(state::ParticleFilterState, new_args::Tuple,
+function pf_update!(state::ParticleFilterView, new_args::Tuple,
                     argdiffs::Tuple, observations::ChoiceMap,
                     proposal::GenerativeFunction, proposal_args::Tuple)
     n_particles = length(state.traces)
@@ -66,10 +63,7 @@ function pf_update!(state::ParticleFilterState, new_args::Tuple,
         end
         state.log_weights[i] += up_weight - prop_weight
     end
-    # Swap references
-    tmp = state.traces
-    state.traces = state.new_traces
-    state.new_traces = tmp
+    update_refs!(state)
     return state
 end
 
@@ -107,7 +101,7 @@ calls to `pf_update!`).
 Similar functionality is provided by [`move_reweight`](@ref), except that
 `pf_update!` also allows model arguments to be updated.
 """
-function pf_update!(state::ParticleFilterState, new_args::Tuple,
+function pf_update!(state::ParticleFilterView, new_args::Tuple,
                     argdiffs::Tuple, observations::ChoiceMap,
                     fwd_proposal::GenerativeFunction, fwd_args::Tuple,
                     bwd_proposal::GenerativeFunction, bwd_args::Tuple)
@@ -122,9 +116,6 @@ function pf_update!(state::ParticleFilterState, new_args::Tuple,
             assess(bwd_proposal, (state.new_traces[i], bwd_args...), discard)
         state.log_weights[i] += up_weight - fwd_weight + bwd_weight
     end
-    # Swap references
-    tmp = state.traces
-    state.traces = state.new_traces
-    state.new_traces = tmp
+    update_refs!(state)
     return state
 end
diff --git a/src/utils.jl b/src/utils.jl
index bd5d6f3..cfab586 100644
--- a/src/utils.jl
+++ b/src/utils.jl
@@ -6,6 +6,17 @@ export mean, var
 using Gen: effective_sample_size
 using Statistics
 
+@inline function update_refs!(state::ParticleFilterState)
+    # Swap references
+    tmp = state.traces
+    state.traces = state.new_traces
+    state.new_traces = tmp
+end
+
+@inline function update_refs!(state::ParticleFilterSubState)
+    state.traces[:] = state.new_traces
+end
+
 lognorm(v::AbstractVector) = v .- logsumexp(v)
 
 """
@@ -14,7 +25,7 @@ lognorm(v::AbstractVector) = v .- logsumexp(v)
 Return the vector of normalized log weights for the current state,
 one for each particle.
 """
-get_log_norm_weights(state::ParticleFilterState) = lognorm(state.log_weights)
+get_log_norm_weights(state::ParticleFilterView) = lognorm(state.log_weights)
 
 """
     get_norm_weights(state::ParticleFilterState)
@@ -22,14 +33,14 @@ get_log_norm_weights(state::ParticleFilterState) = lognorm(state.log_weights)
 Return the vector of normalized weights for the current state,
 one for each particle.
 """
-get_norm_weights(state::ParticleFilterState) = exp.(get_log_norm_weights(state))
+get_norm_weights(state::ParticleFilterView) = exp.(get_log_norm_weights(state))
 
 """
     effective_sample_size(state::ParticleFilterState)
 
 Computes the effective sample size of the particles in the filter.
 """
-Gen.effective_sample_size(state::ParticleFilterState) =
+Gen.effective_sample_size(state::ParticleFilterView) =
     Gen.effective_sample_size(get_log_norm_weights(state))
 
 """
@@ -37,7 +48,7 @@ Gen.effective_sample_size(state::ParticleFilterState) =
 
 Alias for `effective_sample_size`(@ref). Computes the effective sample size.
 """
-get_ess(state::ParticleFilterState) = Gen.effective_sample_size(state)
+get_ess(state::ParticleFilterView) = Gen.effective_sample_size(state)
 
 """
     mean(state::ParticleFilterState[, addr])
@@ -45,10 +56,10 @@ get_ess(state::ParticleFilterState) = Gen.effective_sample_size(state)
 Returns the weighted empirical mean for a particular trace address `addr`.
 If `addr` is not provided, returns the empirical mean of the return value.
 """
-Statistics.mean(state::ParticleFilterState, addr) =
+Statistics.mean(state::ParticleFilterView, addr) =
     sum(get_norm_weights(state) .* getindex.(state.traces, addr))
 
-Statistics.mean(state::ParticleFilterState) =
+Statistics.mean(state::ParticleFilterView) =
     sum(get_norm_weights(state) .* get_retval.(state.traces))
 
 """
@@ -57,10 +68,10 @@ Statistics.mean(state::ParticleFilterState) =
 Returns the empirical variance for a particular trace address `addr`.
 If `addr` is not provided, returns the empirical variance of the return value.
 """
-Statistics.var(state::ParticleFilterState, addr) =
+Statistics.var(state::ParticleFilterView, addr) =
     sum(get_norm_weights(state) .*
         (getindex.(state.traces, addr) .- mean(state, addr)).^2)
 
-Statistics.var(state::ParticleFilterState) =
+Statistics.var(state::ParticleFilterView) =
     sum(get_norm_weights(state) .*
         (get_retval.(state.traces) .- mean(state)).^2)
diff --git a/src/view.jl b/src/view.jl
new file mode 100644
index 0000000..e9eed55
--- /dev/null
+++ b/src/view.jl
@@ -0,0 +1,29 @@
+struct ParticleFilterSubState{U,I,L}
+    traces::SubArray{U,1,Vector{U},I,L}
+    new_traces::SubArray{U,1,Vector{U},I,L}
+    log_weights::SubArray{Float64,1,Vector{Float64},I,L}
+    parents::SubArray{Int,1,Vector{Int},I,L}
+end
+
+Gen.get_traces(state::ParticleFilterSubState) = state.traces
+Gen.get_log_weights(state::ParticleFilterSubState) = state.log_weights
+
+const ParticleFilterView{U} =
+    Union{ParticleFilterState{U}, ParticleFilterSubState{U}} where {U}
+
+function Base.view(state::ParticleFilterState{U},
+                   indices::AbstractVector) where {U}
+    L = Base.viewindexing((indices,)) == IndexLinear()
+    return ParticleFilterSubState{U,Tuple{typeof(indices)},L}(
+        view(state.traces, indices),
+        view(state.new_traces, indices),
+        view(state.log_weights, indices),
+        view(state.parents, indices)
+    )
+end
+
+Base.getindex(state::ParticleFilterState, indices) =
+    Base.view(state, indices)
+
+Base.firstindex(state::ParticleFilterState) = 1
+Base.lastindex(state::ParticleFilterState) = length(state.traces)
diff --git a/test/runtests.jl b/test/runtests.jl
index a46b43d..e708262 100644
--- a/test/runtests.jl
+++ b/test/runtests.jl
@@ -87,6 +87,18 @@ state = pf_update!(state, (10,), (UnknownChange(),), choicemap(),
 @test all([w != 0 for w in get_log_weights(state)])
 end
 
+@testset "Update with different proposals per view" begin
+state = pf_initialize(line_model, (0,), choicemap(), 100)
+substate = pf_update!(state[1:50], (10,), (UnknownChange(),), generate_line(10))
+@test all([tr[:line => 10 => :y] == 0 for tr in get_traces(substate)])
+@test all([w != 0 for w in get_log_weights(substate)])
+substate = pf_update!(state[51:end], (10,), (UnknownChange(),),
+                      generate_line(10), outlier_propose, (10,))
+@test all([tr[:line => 10 => :y] == 0 for tr in get_traces(substate)])
+@test all([tr[:line => 10 => :outlier] == false for tr in get_traces(substate)])
+@test all([w != 0 for w in get_log_weights(state)])
+end
+
 end
 
 @testset "Particle resampling" begin
@@ -240,6 +252,38 @@ rel_weights = parse.(Float64, rel_weights)
 @test all(isapprox.(new_weights, old_weights .+ rel_weights; atol=1e-3))
 end
 
+@testset "Rejuvenation on separate views" begin
+# Log which particles were rejuvenated
+buffer = IOBuffer()
+logger = SimpleLogger(buffer, Logging.Debug)
+state = pf_initialize(line_model, (10,), generate_line(10, 1.), 100)
+old_traces = get_traces(state)[1:50]
+old_weights = get_log_weights(state)[51:end]
+
+with_logger(logger) do
+    pf_move_accept!(state[1:50], metropolis_hastings, (select(:slope),), 1)
+    pf_move_reweight!(state[51:end], move_reweight, (select(:slope),), 1)
+end
+
+# Extract acceptances and relative weights from debug log
+lines = split(String(take!(buffer)), "\n")
+a_lines = filter(s -> occursin("Accepted: ", s), lines)
+accepts = [match(r".*Accepted: (\w+).*", l).captures[1] for l in a_lines]
+accepts = parse.(Bool, accepts)
+r_lines = filter(s -> occursin("Rel. Weight: ", s), lines)
+rel_weights = [match(r".*Rel\. Weight: (.+)\s*", l).captures[1] for l in r_lines]
+rel_weights = parse.(Float64, rel_weights)
+
+# Check that only traces that were accepted are rejuvenated
+new_traces = get_traces(state)[1:50]
+@test all(a ? t1 !== t2 : t1 === t2
+          for (a, t1, t2) in zip(accepts, old_traces, new_traces))
+# Check that weights are adjusted accordingly
+new_weights = get_log_weights(state)[51:end]
+@test all(isapprox.(new_weights, old_weights .+ rel_weights; atol=1e-3))
+
+end
+
 end
 
 @testset "Utility functions" begin