diff --git a/specs/core/0_fork-choice.md b/specs/core/0_fork-choice.md
new file mode 100644
index 000000000..53eccf7ea
--- /dev/null
+++ b/specs/core/0_fork-choice.md
@@ -0,0 +1,96 @@
+# Ethereum 2.0 Phase 0 -- Beacon Chain Fork Choice
+
+**NOTICE**: This document is a work in progress for researchers and implementers.
+
+## Table of contents
+<!-- TOC -->
+
+- [Ethereum 2.0 Phase 0 -- Beacon Chain Fork Choice](#ethereum-20-phase-0----beacon-chain-fork-choice)
+    - [Table of contents](#table-of-contents)
+    - [Introduction](#introduction)
+    - [Constants](#constants)
+        - [Time parameters](#time-parameters)
+    - [Beacon chain processing](#beacon-chain-processing)
+        - [Beacon chain fork choice rule](#beacon-chain-fork-choice-rule)
+
+<!-- /TOC -->
+
+## Introduction
+
+This document represents is the specification for the beacon chain fork choice rule, part of phase 0 of Ethereum 2.0.
+
+## Constants
+
+### Time parameters
+
+| Name | Value | Unit | Duration |
+| - | - | :-: | :-: |
+| `SECONDS_PER_SLOT` | `6` | seconds | 6 seconds |
+
+## Beacon chain processing
+
+Processing the beacon chain is similar to processing the Ethereum 1.0 chain. Clients download and process blocks and maintain a view of what is the current "canonical chain", terminating at the current "head". For a beacon block, `block`, to be processed by a node, the following conditions must be met:
+
+* The parent block with root `block.previous_block_root` has been processed and accepted.
+* An Ethereum 1.0 block pointed to by the `state.latest_eth1_data.block_hash` has been processed and accepted.
+* The node's Unix time is greater than or equal to `state.genesis_time + block.slot * SECONDS_PER_SLOT`.
+
+Note: Leap seconds mean that slots will occasionally last `SECONDS_PER_SLOT + 1` or `SECONDS_PER_SLOT - 1` seconds, possibly several times a year.
+
+Note: Nodes needs to have a clock that is roughly (i.e. within `SECONDS_PER_SLOT` seconds) synchronized with the other nodes.
+
+### Beacon chain fork choice rule
+
+The beacon chain fork choice rule is a hybrid that combines justification and finality with Latest Message Driven (LMD) Greediest Heaviest Observed SubTree (GHOST). At any point in time a [validator](#dfn-validator) `v` subjectively calculates the beacon chain head as follows.
+
+* Abstractly define `Store` as the type of storage object for the chain data and `store` be the set of attestations and blocks that the [validator](#dfn-validator) `v` has observed and verified (in particular, block ancestors must be recursively verified). Attestations not yet included in any chain are still included in `store`.
+* Let `finalized_head` be the finalized block with the highest epoch. (A block `B` is finalized if there is a descendant of `B` in `store` the processing of which sets `B` as finalized.)
+* Let `justified_head` be the descendant of `finalized_head` with the highest epoch that has been justified for at least 1 epoch. (A block `B` is justified if there is a descendant of `B` in `store` the processing of which sets `B` as justified.) If no such descendant exists set `justified_head` to `finalized_head`.
+* Let `get_ancestor(store: Store, block: BeaconBlock, slot: Slot) -> BeaconBlock` be the ancestor of `block` with slot number `slot`. The `get_ancestor` function can be defined recursively as:
+
+```python
+def get_ancestor(store: Store, block: BeaconBlock, slot: Slot) -> BeaconBlock:
+    """
+    Get the ancestor of ``block`` with slot number ``slot``; return ``None`` if not found.
+    """
+    if block.slot == slot:
+        return block
+    elif block.slot < slot:
+        return None
+    else:
+        return get_ancestor(store, store.get_parent(block), slot)
+```
+
+* Let `get_latest_attestation(store: Store, index: ValidatorIndex) -> Attestation` be the attestation with the highest slot number in `store` from the validator with the given `index`. If several such attestations exist, use the one the [validator](#dfn-validator) `v` observed first.
+* Let `get_latest_attestation_target(store: Store, index: ValidatorIndex) -> BeaconBlock` be the target block in the attestation `get_latest_attestation(store, index)`.
+* Let `get_children(store: Store, block: BeaconBlock) -> List[BeaconBlock]` returns the child blocks of the given `block`.
+* Let `justified_head_state` be the resulting `BeaconState` object from processing the chain up to the `justified_head`.
+* The `head` is `lmd_ghost(store, justified_head_state, justified_head)` where the function `lmd_ghost` is defined below. Note that the implementation below is suboptimal; there are implementations that compute the head in time logarithmic in slot count.
+
+```python
+def lmd_ghost(store: Store, start_state: BeaconState, start_block: BeaconBlock) -> BeaconBlock:
+    """
+    Execute the LMD-GHOST algorithm to find the head ``BeaconBlock``.
+    """
+    validators = start_state.validator_registry
+    active_validator_indices = get_active_validator_indices(validators, slot_to_epoch(start_state.slot))
+    attestation_targets = [(i, get_latest_attestation_target(store, i)) for i in active_validator_indices]
+
+    # Use the rounded-balance-with-hysteresis supplied by the protocol for fork
+    # choice voting. This reduces the number of recomputations that need to be
+    # made for optimized implementations that precompute and save data
+    def get_vote_count(block: BeaconBlock) -> int:
+        return sum(
+            start_state.validator_registry[validator_index].high_balance
+            for validator_index, target in attestation_targets
+            if get_ancestor(store, target, block.slot) == block
+        )
+
+    head = start_block
+    while 1:
+        children = get_children(store, head)
+        if len(children) == 0:
+            return head
+        # Ties broken by favoring block with lexicographically higher root
+        head = max(children, key=lambda x: (get_vote_count(x), hash_tree_root(x)))
+```