Categorical Perception

Sounds in language have “phonetic (phonemic) categories,” or simply “phonemes.” These categories are unit of counting sounds in a language. Using those units, we say “There are more than 5 consonants as phonemes in Japanese,Stimulus (S) 1 S 2 S 3 S 4 S 5 S 6 S 7 S 8 S 9 ” “There are about 10 vowels in English,” and so on. 

Once a perceon acquire the phonetic categories of his/her language, it is difficult to change the sound category system. Someimes, especially when you try to learn foreign languages, the fact that the system cannot easily be changed is just bothering and that stands in your way. For example, native speakers of Japanese have difficulties perceiving the difference between the English /r/ and /l/ because these sounds are heard as variants of the Japanese /r/ sound, which is very different from English /r/(Best, 1991; Tomaru et al., 2016, 2017; among others).

Now, can you test whether this person has acquired certain categories, e.g., English /r/ and /l/ categories, thrgouh an experiment?: yes, there are some options. One widely used way is  an experiment of categorical perception can show you whether the person has acqired phonetic categories in question  (Liberman et al., 1957).  Say, we have an (semi ordered) set of sound stimuli, e.g., stimuli 1 through 10, that slightly changes from one phonetic category, e.g., /ba/, to another category, e.g., /da/. Such a set of stimuli is called continuum.  When people hear the continuum, they hear the first half of the stimuli, say stimuli 1 through 5 as good examples of /ba/, and the second half, say stimuli 7 through 10 as good examples of /da/. Stimuli 6 may be the boundary of the identification judgement. At the boundary, people have a chance-level judgements on whether they have heard /ba/ or /da/. Such a boundary is called category (or categoricalbounday.  This suddenchance of perception is called categorical perception.

Figures 1 and 2 are hypothetical (ideal) results of identification and discrimination experiments to be observed when stimuli are categorically perceived. Identification results show an S-curve shape. Discrimination results show peoples’ discrimination judgements are the most accurate when comparing stimuli that cross the categorical boundary observed in the identificaiton data. In other words, people can only be able to discriminate, tell a difference, between presented stimuli when they stimuli belong to different phonetic categories. Using the example above, people can hear the difference between stimuli 5 and 7 because they are at different sides, i.e., categories, of the continuum; however, they can not tell the difference between stimuli 8 and 10 because they are at the same side. When discussing categorical perception, you need both identification and discrimination results.

It is important to note that categorical perception is tend to be observed for perception of native consonants. Therefore, for example, categorical perception of English /r/ and /l/ sounds are observable for native speakers of English, but not for native speakers of Japanese. In addition, it is also worth mentioning that categorical perception is not a definate evidence that people have acquired the sound categories in question. For instance, vowels are less likely to be peceived categorically than consonants are (Eimas, 1963), and sound contexts can change the degree of categorical perception (Tomaru and Arai, 2014, 2016). Moreover, some study reports that non-speech sounds may show categorical-like tendency of perception. You also need to carefully interpet the relationship between the degree of categorical perception and people’s potential ability of language learning. 

Fig. 1: Hypothetical Identification Results
Fig. 2: Hypothetical Discrimination Results

※ Click on [.wav] for high-quality sound.
※ The following /ba/-/da/ continuum can also be found in Demo [K310]
※ The following /ra/-/la/ continuum can also be found Demo [O300]

Continuum
1
Stimulus (S)
1
S 2 S 3 S 4 S 5 S 6 S 7 S 8 S 9
/ba/→/da/
Continuum
2
Stimulus (S)
1
S 2 S 3 S 4 S 5 S 6 S 7 S 8 S 9
/ra/→/la/

<Demo-experiment of Categorical Perception>

(1) Identification Task:

a. Continuum 1
・Which do you hear?: /ba/ or /da/

Trial 1 Trial 2 Trial 3

*See answers

b. Continuum 2
・Which do you hear?: /ra/ or /la/

Trial 1 Trial 2 Trial 3

*See answers

(2) AXB Discrimination Task:

a. Continuum 1
・You will hear 3 stimuli in sequence on each trial. Say whether  the SECOND stimulus is more similar to  the first or the third.

Trial 1 Trial 2 Trial 3

*See answers

b. Continuum 2
・You will hear 3 stimuli in sequence on each trial. Say whether  the SECOND stimulus is more similar to  the first or the third.

Trial 1 Trial 2 Trial 3

*See answers

  1. Best, C. T, The emergence of native-language phonological influences in infants: a perceptual assimilation model, Haskins Laboratories Status Report on Speech Research, SR-107/108, 1-30, 1991.
  2. Eimas, P. D., The relation between identification and discrimination along speech and non-speech continua, Language and Speech, 206-217, 1963.
  3. Liberman, A. M., Harris, K. S., Hoffman, H. S., and Griffith, B. C., Journal of Experimental Psychology, 54 (5), 358-368, 1957.
  4. Tomaru, K., and Arai, T., Discrimination of /ra/-/la/ speech continuum by native speakers of English under nonisolated conditions, Acoustical Science and Technology, 251-259, 2014.
  5. Tomaru, K., and Arai, T., Role of labeling mediation in speech perception: Evidence from a voiced stop continuum perceived in different surrounding sound contexts, 303-314, 2016.
  6. Tomaru, K., Nakamura, T., and Arai, T., Perceptual vowel insertion to the pre-syllable positions of English /ra/ and /la/ syllables by native speakers of Japanese, Proceedings of the Spring Meeting of Acoustical Society of Japan, 441-444, 2016.
  7. Tomaru, K., Nakamura, T., and Arai, T., Perception of English /r/ and /l/ by native speakers of Japanese under the condition of onset lengthening, Proceedings of the Spring Meeting of Acoustical Society of Japan, 1483-1484, 2017.