diff --git a/onihime/src/lexer/error.rs b/onihime/src/lexer/error.rs
new file mode 100644
index 0000000..58e6d7c
--- /dev/null
+++ b/onihime/src/lexer/error.rs
@@ -0,0 +1,52 @@
+use crate::span::Span;
+
+/// Errors that can occur during lexical analysis.
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub enum LexerErrorKind {
+    /// An invalid character was encountered.
+    InvalidChar,
+    /// An invalid escape sequence was encountered.
+    InvalidEscape(char),
+    /// An invalid number was encountered.
+    InvalidNumber(String),
+    /// An invalid string was encountered.
+    InvalidString,
+    /// An unclosed character literal was encountered.
+    UnclosedChar,
+    /// And unclosed string literal was encountered.
+    UnclosedString,
+}
+
+/// Lexer error, with a start and end location.
+#[derive(Debug, Clone, PartialEq)]
+pub struct LexerError {
+    /// The type of error encountered.
+    pub kind: LexerErrorKind,
+    /// The span in which the error occurred.
+    pub span: Span,
+}
+
+impl LexerError {
+    /// Construct a new instance of `LexerError`.
+    #[must_use]
+    pub const fn new(kind: LexerErrorKind, span: Span) -> Self {
+        Self { kind, span }
+    }
+}
+
+impl std::error::Error for LexerError {}
+
+impl std::fmt::Display for LexerError {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        use LexerErrorKind::*;
+
+        match &self.kind {
+            InvalidChar => write!(f, "Invalid character literal"),
+            InvalidEscape(c) => write!(f, "Unknown escape sequence '\\{c}' in string"),
+            InvalidNumber(n) => write!(f, "`{n}` is not a valid numeric literal"),
+            InvalidString => write!(f, "Invalid string literal"),
+            UnclosedChar => write!(f, "Unclosed character literal"),
+            UnclosedString => write!(f, "Unclosed string literal"),
+        }
+    }
+}
diff --git a/onihime/src/lexer/mod.rs b/onihime/src/lexer/mod.rs
new file mode 100644
index 0000000..b065461
--- /dev/null
+++ b/onihime/src/lexer/mod.rs
@@ -0,0 +1,399 @@
+use std::{
+    str::{Chars, FromStr},
+    sync::Arc,
+};
+
+pub use self::{
+    error::{LexerError, LexerErrorKind},
+    symbol::Symbol,
+    token::{Token, TokenKind},
+};
+use crate::span::{Source, Span};
+
+mod error;
+mod symbol;
+mod token;
+
+/// A trait for checking if a character is a separator.
+pub trait Separator {
+    /// Check if the character is a separator.
+    fn is_separator(&self) -> bool;
+}
+
+impl Separator for char {
+    fn is_separator(&self) -> bool {
+        self.is_ascii_whitespace() || matches!(self, '(' | ')' | '[' | ']' | '{' | '}' | ',')
+    }
+}
+
+/// A lexer, used by the parser.
+#[derive(Debug)]
+pub struct Lexer<'lexer> {
+    input: Chars<'lexer>,
+    byte: usize,
+    source: Arc<Source>,
+}
+
+impl<'lexer> Lexer<'lexer> {
+    /// Create a new lexer instance from a string.
+    #[must_use]
+    pub fn new(input: &'lexer str) -> Self {
+        let source = Arc::new(Source::new(None, input.to_string()));
+
+        Self {
+            input: input.chars(),
+            byte: 0,
+            source,
+        }
+    }
+
+    /// Set the name of the [Source] being lexically analyzed.
+    pub fn set_name(&mut self, name: String) {
+        if let Some(source) = Arc::get_mut(&mut self.source) {
+            source.set_name(name);
+        } else {
+            unimplemented!(); // FIXME: What should we do in this case?
+        }
+    }
+
+    /// The source being lexically analyzed.
+    #[must_use]
+    pub fn source(&self) -> Arc<Source> {
+        self.source.clone()
+    }
+
+    /// Get the current character.
+    #[must_use]
+    pub fn current(&self) -> Option<char> {
+        self.input.as_str().chars().next()
+    }
+
+    /// Get the unparsed input.
+    #[must_use]
+    pub fn get_unparsed(&self) -> &str {
+        self.input.as_str()
+    }
+
+    /// Get the current position of the lexer.
+    #[inline]
+    #[must_use]
+    pub(crate) fn span(&self) -> Span {
+        Span::new(self.byte..self.byte, self.source.clone())
+    }
+
+    /// Get the nth character ahead of the current character without advancing.
+    fn peek(&self, n: usize) -> Option<char> {
+        self.input.as_str().chars().nth(n)
+    }
+
+    /// Advance the lexer by one character.
+    fn advance(&mut self) -> Option<char> {
+        let c = self.input.next()?;
+        self.byte += c.len_utf8();
+
+        Some(c)
+    }
+
+    /// Read a word from the input until a separator is reached.
+    fn read_word(&mut self) -> String {
+        let mut word = String::new();
+        while let Some(c) = self.current() {
+            if c.is_separator() {
+                break;
+            }
+
+            word.push(c);
+            self.advance();
+        }
+
+        word
+    }
+
+    /// Parse a value from the input or return an error.
+    fn parse_or<T>(&mut self, err: impl Fn(String) -> LexerErrorKind) -> Result<T, LexerError>
+    where
+        T: FromStr,
+    {
+        let span = self.span();
+        let word = self.read_word();
+
+        word.parse()
+            .map_err(|_| LexerError::new(err(word), span.join(&self.span())))
+    }
+
+    /// Read the next token from the input.
+    pub fn read(&mut self) -> Result<Option<Token>, LexerError> {
+        let c = loop {
+            match self.current() {
+                Some(c) if c.is_ascii_whitespace() || c == ',' => {
+                    self.advance();
+                }
+                Some(c) => break c,
+                None => return Ok(None),
+            }
+        };
+
+        let mut span = self.span();
+        let kind = match c {
+            '(' => {
+                self.advance();
+                TokenKind::OpenParen
+            }
+            ')' => {
+                self.advance();
+                TokenKind::CloseParen
+            }
+            '{' => {
+                self.advance();
+                TokenKind::OpenBrace
+            }
+            '}' => {
+                self.advance();
+                TokenKind::CloseBrace
+            }
+            '[' => {
+                self.advance();
+                TokenKind::OpenBracket
+            }
+            ']' => {
+                self.advance();
+                TokenKind::CloseBracket
+            }
+            '0'..='9' => TokenKind::Number(self.parse_or(LexerErrorKind::InvalidNumber)?),
+            '+' | '-' if matches!(self.peek(1), Some('0'..='9')) => {
+                TokenKind::Number(self.parse_or(LexerErrorKind::InvalidNumber)?)
+            }
+            ';' => {
+                let mut comment = String::new();
+                while let Some(c) = self.advance() {
+                    match c {
+                        ';' => continue,
+                        '\n' => break,
+                        c => {
+                            comment.push(c);
+                        }
+                    }
+                }
+
+                TokenKind::LineComment(comment.trim().into())
+            }
+            '#' if matches!(self.peek(1), Some('|')) => {
+                self.advance(); // '#'
+                self.advance(); // '|#'
+
+                let mut comment = String::new();
+                while let Some(c) = self.advance() {
+                    match c {
+                        '|' if matches!(self.peek(0), Some('#')) => {
+                            self.advance(); // '|'
+                            self.advance(); // '#'
+                            break;
+                        }
+                        c => {
+                            comment.push(c);
+                        }
+                    }
+                }
+
+                TokenKind::BlockComment(comment.trim().into())
+            }
+            ':' => {
+                self.advance();
+                TokenKind::Keyword(Symbol(self.read_word()))
+            }
+            '"' => {
+                self.advance(); // '"'
+
+                let quote_span = span.clone().join(&self.span());
+                let mut string = String::new();
+
+                loop {
+                    let ch_span = self.span();
+                    string.push(match self.advance() {
+                        Some('"') => break,
+                        Some('\\') => match self.advance() {
+                            Some(c @ ('"' | '\\')) => c,
+                            Some('n') => '\n',
+                            Some('e') => '\x1b',
+                            Some(c) => {
+                                return Err(LexerError::new(
+                                    LexerErrorKind::InvalidEscape(c),
+                                    ch_span.join(&self.span()),
+                                ))
+                            }
+                            None => {
+                                return Err(LexerError::new(
+                                    LexerErrorKind::UnclosedString,
+                                    quote_span,
+                                ))
+                            }
+                        },
+                        Some(c) => c,
+                        None => {
+                            return Err(LexerError::new(LexerErrorKind::UnclosedString, quote_span))
+                        }
+                    });
+                }
+
+                if self.current().is_some_and(|c| !c.is_separator()) {
+                    self.read_word();
+                    return Err(LexerError::new(
+                        LexerErrorKind::InvalidString,
+                        span.join(&self.span()),
+                    ));
+                }
+
+                TokenKind::String(string)
+            }
+            '\'' => {
+                self.advance(); // '\''
+
+                let c = match self.advance() {
+                    Some('\\') => match self.advance() {
+                        Some(c @ ('"' | '\\')) => c,
+                        Some('n') => '\n',
+                        Some('e') => '\x1b',
+                        Some(c) => {
+                            return Err(LexerError::new(
+                                LexerErrorKind::InvalidEscape(c),
+                                span.join(&self.span()),
+                            ));
+                        }
+                        None => {
+                            return Err(LexerError::new(
+                                LexerErrorKind::UnclosedChar,
+                                span.join(&self.span()),
+                            ));
+                        }
+                    },
+                    Some(c) => c,
+                    None => {
+                        return Err(LexerError::new(
+                            LexerErrorKind::UnclosedChar,
+                            span.join(&self.span()),
+                        ))
+                    }
+                };
+
+                if self.advance() != Some('\'') {
+                    self.read_word();
+                    return Err(LexerError::new(
+                        LexerErrorKind::InvalidChar,
+                        span.join(&self.span()),
+                    ));
+                }
+
+                TokenKind::Char(c)
+            }
+            _ => {
+                let word = self.read_word();
+                match word.as_str() {
+                    "true" => TokenKind::Bool(true),
+                    "false" => TokenKind::Bool(false),
+                    "nil" => TokenKind::Nil,
+                    _ => TokenKind::Symbol(Symbol::from(word)),
+                }
+            }
+        };
+
+        span.extend(&self.span());
+
+        Ok(Some(Token::new(kind, span)))
+    }
+}
+
+impl Iterator for Lexer<'_> {
+    type Item = Result<Token, LexerError>;
+
+    fn next(&mut self) -> Option<Self::Item> {
+        self.read().transpose()
+    }
+}
+
+#[cfg(test)]
+mod tests {
+    use super::*;
+
+    macro_rules! test {
+        ( $name:ident: $input:literal, $tokens:expr ) => {
+            #[test]
+            fn $name() {
+                let mut lexer = Lexer::new($input);
+
+                for token in $tokens {
+                    let x = lexer.next().map(|r| match r {
+                        Ok(t) => Ok(t.kind),
+                        Err(e) => Err(e.kind),
+                    });
+
+                    assert_eq!(x, Some(token));
+                }
+
+                assert_eq!(lexer.next(), None);
+            }
+        };
+    }
+
+    test!(lex: "(+ 14 25.5 333 (* 2 5))", [
+        Ok(TokenKind::OpenParen),
+        Ok(TokenKind::Symbol(Symbol::from("+"))),
+        Ok(TokenKind::Number(14.0)),
+        Ok(TokenKind::Number(25.5)),
+        Ok(TokenKind::Number(333.0)),
+        Ok(TokenKind::OpenParen),
+        Ok(TokenKind::Symbol(Symbol::from("*"))),
+        Ok(TokenKind::Number(2.0)),
+        Ok(TokenKind::Number(5.0)),
+        Ok(TokenKind::CloseParen),
+        Ok(TokenKind::CloseParen),
+    ]);
+
+    test!(newline: "(+ 14 25.5 333\n(* 2 5 5.x))", [
+        Ok(TokenKind::OpenParen),
+        Ok(TokenKind::Symbol(Symbol::from("+"))),
+        Ok(TokenKind::Number(14.0)),
+        Ok(TokenKind::Number(25.5)),
+        Ok(TokenKind::Number(333.0)),
+        Ok(TokenKind::OpenParen),
+        Ok(TokenKind::Symbol(Symbol::from("*"))),
+        Ok(TokenKind::Number(2.0)),
+        Ok(TokenKind::Number(5.0)),
+        Err(LexerErrorKind::InvalidNumber("5.x".into())),
+        Ok(TokenKind::CloseParen),
+        Ok(TokenKind::CloseParen),
+    ]);
+
+    test!(negative_minus: "(- 1 -2 3)", [
+        Ok(TokenKind::OpenParen),
+        Ok(TokenKind::Symbol(Symbol::from("-"))),
+        Ok(TokenKind::Number(1.0)),
+        Ok(TokenKind::Number(-2.0)),
+        Ok(TokenKind::Number(3.0)),
+        Ok(TokenKind::CloseParen),
+    ]);
+
+    test!(line_comment: "; foo\n;; bar baz qux", [
+        Ok(TokenKind::LineComment("foo".into())),
+        Ok(TokenKind::LineComment("bar baz qux".into())),
+    ]);
+
+    test!(block_comment: "#| foo\nbar |#", [
+        Ok(TokenKind::BlockComment("foo\nbar".into()))
+    ]);
+
+    test!(error_parse_numbers: "2 55 3.144 0.0001 1.1.1", [
+        Ok(TokenKind::Number(2.0)),
+        Ok(TokenKind::Number(55.0)),
+        Ok(TokenKind::Number(3.144)),
+        Ok(TokenKind::Number(0.0001)),
+        Err(LexerErrorKind::InvalidNumber("1.1.1".into())),
+    ]);
+
+    test!(error_unclosed_string: "\"hiii", [
+        Err(LexerErrorKind::UnclosedString),
+    ]);
+
+    test!(error_invalid_string: "\"hiii\"222", [
+        Err(LexerErrorKind::InvalidString),
+    ]);
+}
diff --git a/onihime/src/lexer/symbol.rs b/onihime/src/lexer/symbol.rs
new file mode 100644
index 0000000..3f7c697
--- /dev/null
+++ b/onihime/src/lexer/symbol.rs
@@ -0,0 +1,19 @@
+/// A symbol used to identify a function or variable.
+#[derive(Debug, Clone, PartialEq, Eq, PartialOrd, Ord, Hash)]
+pub struct Symbol(pub String);
+
+impl Symbol {
+    /// Create a new `Symbol` from a string.
+    pub fn from<S>(s: S) -> Self
+    where
+        S: Into<String>,
+    {
+        Self(s.into())
+    }
+}
+
+impl std::fmt::Display for Symbol {
+    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
+        write!(f, "{}", self.0)
+    }
+}
diff --git a/onihime/src/lexer/token.rs b/onihime/src/lexer/token.rs
new file mode 100644
index 0000000..f3f9c6f
--- /dev/null
+++ b/onihime/src/lexer/token.rs
@@ -0,0 +1,56 @@
+use super::Symbol;
+use crate::span::Span;
+
+/// The type of a [Token].
+#[derive(Debug, Clone, PartialEq)]
+pub enum TokenKind {
+    /// Opening parenthesis, e.g. '('
+    OpenParen,
+    /// Closing parenthesis, e.g. ')'
+    CloseParen,
+    /// Opening brace, e.g. '{'
+    OpenBrace,
+    /// Closing brace, e.g. '}'
+    CloseBrace,
+    /// Opening bracket, e.g. '['
+    OpenBracket,
+    /// Closing bracket, e.g. ']'
+    CloseBracket,
+
+    /// Block comment, e.g. '#| ... |#'
+    BlockComment(String),
+    /// Line comment, e.g. '; ...'
+    LineComment(String),
+
+    /// Boolean, e.g. 'true', 'false'
+    Bool(bool),
+    /// Character, e.g. 'c', '\n'
+    Char(char),
+    /// Number, e.g. '1', '2.0', '0.003'
+    Number(f64),
+    /// String, e.g. '"foo bar"'
+    String(String),
+    /// Keyword, e.g. ':baz'
+    Keyword(Symbol),
+    /// Symbol, e.g. 'qux', '+'
+    Symbol(Symbol),
+    /// Nil, e.g. 'nil'
+    Nil,
+}
+
+/// A token with a start and end location.
+#[derive(Debug, Clone, PartialEq)]
+pub struct Token {
+    /// The type of token.
+    pub kind: TokenKind,
+    /// The span in which the token occurs.
+    pub span: Span,
+}
+
+impl Token {
+    /// Construct a new instance of `Token`.
+    #[must_use]
+    pub const fn new(kind: TokenKind, span: Span) -> Self {
+        Self { kind, span }
+    }
+}
diff --git a/onihime/src/lib.rs b/onihime/src/lib.rs
index 0cffc7e..b5ef643 100644
--- a/onihime/src/lib.rs
+++ b/onihime/src/lib.rs
@@ -2,4 +2,5 @@
 
 #![deny(missing_debug_implementations, missing_docs, rust_2018_idioms)]
 
+mod lexer;
 mod span;