Page content

The empty set approach compares two sounds that are unknown to the child (not used - e.g.  /l/ vs /tʃ/) in non-homonymous contrasts. If the child’s phonetic inventory consists of: [m, n, t, d, p, b, w, j, h], /l/ and /tʃ/ can be used as targets because they differ in place (alveolar vs postalveolar), manner (approximant vs affricate) and voicing (voiced vs voiceless).

McLeod and Baker (2017, p.444)

Suitable children

  • Consistent phonological impairment
  • Severity: moderate - severe
  • Age: between 2;08 - 7;11 years, with most children being around 4 years
  • Exclusion of at least 5 - 7 sounds from their phonetic/phonemic inventories

(Williams et al. 2010, Gierut 1989, 2008)

Empty Set procedure

Adapted from McLeod and Baker 2017, p.444-446; Williams et al. 2010, p.109

  1. Two sounds that are not present in the child's phonemic inventory are targeted.
  2. The target sounds in initial position are used to create 8 real or non-word pairs.
  3. Storkel (2018b) provides evidence-based guidance on how to best promote generalisation through the vocabulary characteristics chosen for the target words: high frequency + high density; low frequency + high density; high frequency + mixed density; low frequency + later acquired; or nonwords. Note that density refers to the number of words that sounds similar to one another, with words from high density neighbourhoods having many similarly sounding words (i.e., bat, fat, rat, mat, sat.......) compared to words from low density neighbourhoods which do not (and are also often less frequent words) (i.e., umbrella).
  4. If non-words are used, the child is familiarised with them and they are assigned meaning through storytelling (i.e., via aliens, mermaids, objects or actions etc.). If real words are used, the child is also familiarised with these.
  5. Imitation phase: the child imitates the SLTs' model for each real/non-word. The child is praised for correct imitations and given instructional feedback for incorrect imitations. Once the child achieves 75% accuracy across 2 consecutive sessions or completes 7 sessions, they can move on to the next therapy stage.
  6. Spontaneous phase: the child produces the target real/non-words without a model and is praised throughout attempts. Once the child achieves 90% accuracy over 3 sessions, or when 12 sessions have been completed, the child can progress to the next stage of therapy. SLTs should use their own clinical judgement to set the child's performance criteria (e.g. 50%) in conversational speech.
  7. Generalisation probe: A generalisation probe (i.e., using non-treated words/simple phrases which include the target in the treated position) is administered. An accuracy level of approximately 70% in conversational speech probes is aimed for, for therapy on this target to be terminated.
  8. At this point there are two options: (1) another therapy target may be chosen or; (2) the child may be discharged. SLTs should use their own clinical judgement here.

Note: Clinical judgement should be used alongside the child's assessment data when choosing and implementing an intervention protocol.


  • Probes should be developed to monitor and detect changes in a child’s speech sound system.
  • Probes should consider within and across class generalisation, taking into consideration the implicational changes expected (see the target selection section for an overview).
  • Probes should consider generalisation across word positions and generalisation from single words to conversational speech.
  • See Gierut (2008a) for further information.

Monitoring progress

  • If the child is not progressing to the accuracy levels expected within the time-frame expected (as identified in the procedure), then a review may be necessary. SLTs must use their clinical judgement here, as well as the pre-determined criteria set for each individual child.

Supporting literature


Intervention intensity

Find out more information on the intervention intensity page about the empty set approach.