Orthographic Processing in Baboons (Papio papio)

See allHide authors and affiliations

Science  13 Apr 2012:
Vol. 336, Issue 6078, pp. 245-248
DOI: 10.1126/science.1218152

You are currently viewing the figures only.

View Full Text

Log in to view the full text

Log in through your institution

Log in through your institution

  1. Fig. 1

    Teaching baboons to recognize words. (A) Skilled readers use an orthographic code to recognize words, mapping elementary visual features, such as lines of different orientation (here features contained in the word “WASP”), onto whole-word orthographic representations via some form of letter-level code (14, 9). (B and C) While maintained in their social group, the baboons had free access to computer-controlled operant conditioning setups with touch screen technology (13). (C) The baboons were trained to recognize four-letter English words and distinguish them from strings of letters that are not English words, such as “STOD.” Baboons responded by touching either the cross or the oval shape presented immediately after the word or nonword. After a correct response, a blank screen was presented and baboons received a food reward (dry wheat). A green screen was presented for 3 s after an incorrect response. We asked whether baboons would use an orthographic code, as described in (A), in order to discriminate words from nonwords.

  2. Fig. 2

    Successful word-nonword discrimination in baboons. (A) Accuracy for words (e.g., DONE, LAND, THEM, VAST) and (B) nonwords (e.g., DRAN, LONS, TELK, VIRT) was calculated for blocks of 2000 consecutive trials (except for the last block) separately for each baboon (here and in Figs. 3 and 4, baboons are indicated by their abbreviated names: DAN, ART, CAU, DOR, VIO, and ARI). The results of a signal detection analysis are shown in (C) (sensitivity: baboons’ ability to discriminate words from nonwords) and (D) (bias: baboons’ inclination to answer “word” or “nonword”). During the first block of 2000 trials, numerical estimates of bias show that each baboon predominantly chose one of the two possible responses resulting in a “word” or “nonword” bias and low sensitivity. After 2000 trials, the baboons started to perform accurate word-nonword classification by responding “word” to repeated stimuli and “nonword” to novel stimuli as shown by above-zero sensitivities and bias values close to zero. Baboons attained an accuracy level of about 75%. Error bars in (A) and (B) correspond to the 95% binomial confidence interval, which are also displayed in gray for chance performance (see supplementary materials for more details).

  3. Fig. 3

    Percentage of nonword responses on trials corresponding to words seen for the first time as compared to the first nonword stimuli after these particular trials. Performance on trials corresponding to the first presentation of words is of particular interest, because any divergence from performance to nonword stimuli is an indication that the baboons have learned general statistical properties of the two classes of stimuli. All six baboons showed such a divergence for both the total number of first word trials (A) and the last 50 first word trials (B), as revealed in the differences in the percentage of nonword responses to first words and nonwords (all P values < 0.01).

  4. Fig. 4

    Performance in response to nonwords depends on their orthographic similarity to learned words for both monkeys and humans. (A) For each of the last 20,000 nonword trials, the orthographic Levenshtein distance (OLD20) (15, 16) was computed between the corresponding nonword and each of the words learned at that time, separately for each baboon. The average accuracy corresponding to each unique value of OLD20 was then calculated. The graph shows that baboons responded less accurately to more wordlike nonwords (smaller OLD20 values). Errors bars correspond to the 95% binomial confidence interval. (B) For comparison, humans show a similar sensitivity to orthographic distance to known words when responding to nonwords (see supplementary text).