<p>
The overall structure of the sentence (including its ambiguity)
is expressed by feature structure F1:
<xmp>
    F1:  Analysis of sentence <q>Wash sinks.</q>
    Category = Sentence
    Alternative subanalyses = /F1a/ OR /F1b/
</xmp>
<!>
The first subanalysis is F1a:
<xmp>
    F1a:  One subanalysis of <q>Wash sinks.</q>
    Mood = Indicative
    Voice = Stative
    Subject = /F3/
    Predicate = /F6/
</xmp>
<!>
The second analysis is F1b:
<xmp>
    F1b:  Another subanalysis of <q>Wash sinks.</q>
    Mood = Imperative
    Voice = Active
    Subject =
                   Category = NP            /F1b.F1/
                   Number = Unspecified
                   Person = 2
 
    Predicate = /F2/
</xmp>
<!>
The verb phrase <q>wash sinks</q> is analyzed in
feature structure F2.  Note that the ambiguity of the verb phrase,
which resides in the lexical ambiguity of the direct object, is not
represented here, but in the feature structure describing the direct
object.  The lexical ambiguity of the verb <q>wash</q> is represented
within feature structure F8 (pointed at here as the head of this
verb phrase); we indicate with an attribute <term>path</term> or
on the <tag>f.ptr</tag> tag which sub-structures in the target structure
are selected in the context of the analysis.
    <fn>Alternate name:  <term>qualifier</term>.</fn>
The value of the <term>path</term> attribute is a list of feature
structure IDs which apply to the word occurrence in question.
<term>Path</term> has meaning only when pointing at feature
structures contained in an <tag>f.s.OR</tag> tag.  If the feature
structure pointed at by the <term>target</term> attribute contains more
than one <tag>f.s.OR</tag>, then the <term>path</term> attribute for
the pointer may contain pointers to children of each <tag>f.s.OR</tag>
in the target feature structure.
If the ambiguity represented by an <tag>f.s.OR</tag> is to be
preserved, then the <term>path</term> value can contain more than one
child of the <tag>f.s.OR</tag>.
The ID of the
<tag>f.s.OR</tag> itself can be specified, as a shorthand for the
IDs of all its children.
If the <term>path</term> value contains no IDs for any feature structure
contained in a given <tag>f.s.OR</tag>, then the implication is that all
are selected.
<note>Possible conflict here with general principles of underspecified
analyses as described in EDW12.  -Ed.</note>
<xmp>
    F2:  Analysis of VP <q>wash sinks</q>
    Category = VP
    Head = /F8/ (path = /F8.i/)
    (F8 is ambiguous -- we choose branch F8.i within the choice F8.or)
    Direct Object = /F5/
</xmp>
<!>
The (ambiguous) noun phrase <q>Wash</q> is analyzed in
feature structure F3:
<xmp>
    F3:  Analysis of NP <q>Wash</q>
    Category = NP
    Alternative subanalyses of NP <q>Wash</q> [F3.or] =
              +PROPER                  [F3.proper]
              Number = Singular
              Object-type = Person
 
              OR
 
              +MASS                    [F3.mass]
              Number = Singular
              Head = /F7/
</xmp>
<!>
The verb phrase <q>wash</q> required by the imperative analyses
is analyzed in
feature structure F4.
<xmp>
    F4:  Analysis of VP <q>wash</q>
    Category = VP
    Head = /F8/ (path = /F8.intr/
</xmp>
<!>
The noun phrase <q>sinks</q> is analyzed in
feature structure F5.  Note that the lexical ambiguity of the noun
itself is not directly reflected in this description of the phrase it
heads.
<xmp>
    F5:  Analysis of NP <q>sinks</q>
    Category = NP
    +COUNT
    Number = Plural
    Head = /F9/
</xmp>
<!>
The verb phrase <q>sinks</q> is analyzed in
feature structure F6.
<xmp>
    F6:  Analysis of VP <q>sinks</q>
    Category = VP
    Head = /F10/ (path = /F10.intr/)
</xmp>
<!>
The common noun <q>wash</q> is analyzed in
feature structure F7.
<xmp>
    F7:  Analysis of noun <q>wash</q>
    Category = Noun
    -INFLECTED
</xmp>
<!>
The word <q>wash</q> is analyzed as a verb in
feature structure F8.  Note that both transitive and intransitive
are specified for the verb; an analysis constructed exclusively for
the example sentence might not include the transitive feature, but
a general analysis contained for example in a lexicon would have to
include both characterizations.  The pointers to the feature from
other structures in the analysis must therefore choose between the
two possibilities.
<xmp>
    F8:  Analysis of verb <q>wash</q>
    Category = Verb
    -INFLECTED
    Alternative subanalyses of verb <q>wash</q> [F8.or]
 
         F8.intr:
         Subcat = Intransitive
 
         OR
 
         F8.tran:
         Subcat = Transitive
</xmp>
<!>
The lexically ambiguous noun <q>sinks</q> is analyzed in
feature structure F9.  The lexical ambiguity is expressed as an
alternation between two glosses.
<xmp>
    F9:  Analysis of noun <q>sinks</q>
    Category = Noun
    Gloss = 'basins for washing' OR 'depressions in a land surface'
    +INFLECTED
    Number = Plural
</xmp>
<!>
The verb <q>sinks</q> is analyzed in
feature structure F10.  Like <q>wash</q>, this verb can be
either transitive or intransitive, and other analyses may disambiguate
the word by indicating a preferred alternative.
<xmp>
    F10:  Analysis of verb <q>sinks</q>
    Category = Verb
    +INFLECTED
    Number of Subject = Singular
    Person of Subject = 3
    Alternative subanalyses of verb <q>sinks</q> [F10.or]:
              F10.intr:
         Subcat = Intransitive
 
         OR
 
         F10.tran:
         Subcat = Transitive
</xmp>
