Neuroscience Letters 316 (2001) 71–74

www.elsevier.com/locate/neulet

Getting it: human event-related brain response to jokes in

good and poor comprehenders
Seana Coulson*, Marta Kutas
University of California, San Diego, CA, USA
Received 13 August 2001; received in revised form 8 October 2001; accepted 8 October 2001

Abstract
Joke comprehension has been decomposed into surprise registration followed by a coherence stage, involving frame-
shifting (retrieving a new frame from long-term memory to reinterpret information in working memory). We examined
this view by recording event-related brain potentials (ERPs) from adults reading one-line jokes or non-joke controls with
equally unexpected endings. Joke and non-joke ERPs differed in several respects depending on participants’ ability to
get the joke and contextual constraint. In good joke comprehenders, all jokes elicited a left-lateralized sustained nega-
tivity (500–900 ms), indexing frame-shifting, low constraint jokes elicited a frontal positivity (500–900 ms), and high
constraint jokes elicited an N400 and later posterior positivity. By contrast, poor joke comprehenders showed only a right
frontal negativity (300–700 ms) to jokes. This pattern of effects does not map readily onto a two-stage model of joke
comprehension. q 2001 Published by Elsevier Science Ireland Ltd.
Keywords: Event-related potentials; Comprehension; Jokes; Language; Humor

The ability to appreciate humor is an intriguing aspect of patients with right hemisphere brain damage (RHD), espe-
human behavior, considered by many to be a defining cially including the anterior portion of the right frontal lobe
human attribute [11]. Though it recruits a number of cogni- [3,14]. When asked to pick the punch-line of a joke from an
tive processes, analysts have decomposed joke comprehen- array of choices, including straightforward endings, non-
sion into two major components: registration of surprise sequitur endings, and the correct punch-line, RHD patients
followed by re-establishment of coherence [16]. For exam- erred by picking non-sequitur endings, indicating that they
ple, ‘years’ is surprising when it occurs in “I let my accoun- know surprise is necessary but are impaired on coherence
tant do my taxes because it saves time: last spring it saved [2]. Though these data suggest a dissociation between the
me 10 years”. However, to really ‘get’ the joke, the listener surprise and coherence stages of joke comprehension, their
must go beyond surprise and formulate a new, coherent implications for normal brain function are uncertain, as they
interpretation in which the speaker is worried about going may reflect compensatory strategies, and functional reorga-
to jail, and pays an accountant to conceal illegal business nization of the damaged brain. Further, lesion data do not
practices. Coherence involves a process of frame-shifting, in address the time course of the two stages of joke compre-
which the listener activates a new frame from long-term hension. To these ends, we recorded ERPs as neurologically
memory to reinterpret information already active in working intact participants read sentences that ended either as jokes
memory [4]. Here ,we use event-related brain potentials or with equally surprising non-joke endings that did not
(ERPs) recorded from healthy adults to examine the two- entail frame-shifting.
stage model by assessing the psychological reality of frame- Participants were 28 right-handed, college-age, monolin-
shifting, and differentiate it from the surprise component of gual English speakers (ten men) with normal, or corrected-
joke comprehension. to-normal, vision. Participants’ read sentences and
Joke comprehension deficits have been observed in answered true/false questions while their electroencephalo-
gram was sampled at 250 Hz and recorded from 26 tin
electrodes arranged in a geodesic pattern, referenced to
* Corresponding author. Cognitive Science 0515, 9500 Gilman the left mastoid.
Drive, La Jolla, CA 92093-0515, USA. Tel.: 11-858-822-4037; fax:
11-858-534-1228.
Sixty sentences ended either as jokes (30), or as non-jokes
E-mail address: [email protected] (S. Coulson). (30); 100 were filler sentences. One-line jokes were chosen
0304-3940/01/$ - see front matter q 2001 Published by Elsevier Science Ireland Ltd.
PII: S03 04 - 394 0( 0 1) 02 38 7- 4
72 S. Coulson, M. Kutas / Neuroscience Letters 316 (2001) 71–74

not participants were getting the jokes. For example, for “I

asked the bartender for something cold and full of rum, and
he recommended his wife”, the comprehension question
was “The bartender’s wife was a frigid lush” (true). For
“We have a concrete swimming pool, but when you live
near the beach, nobody wants to swim in the concrete”,
the question was “Their pool is filled with chlorinated
water” (false).
ERPs to sentence-final words displayed the P1/N1/P2
complex typical of visually presented words, a negative-
going wave from 300 to 700 ms (N400), and a late positivity
(see Fig. 1). ERPs were quantified as the mean amplitude
between 300–500, 500–700, and 700–900 ms post-word
onset relative to 100 ms pre-stimulus baseline and subjected
to repeated measures analysis of variance (ANOVA) with
ending (joke/non-joke), constraint (low/high), and three
factors indexing scalp topography: hemisphere (left/right);
laterality (dorsal/lateral); and anterior/posterior (four levels).
All P values were corrected using the Huhyn–Feldt correc-
tion; for clarity, we report original degrees of freedom.
Between 300 and 500 ms, there was a main effect of
ending (Fð1;27Þ ¼ 7:98, P , 0:01) and an ending by
constraint interaction (Ending £ Constraint, Fð1;27Þ ¼ 4:34,
Fig. 1. Grand average (n ¼ 28) ERPs to joke (dotted) and non-joke
(solid) endings from all electrode (n ¼ 26) sites. Negative voltage
P , 0:05; Ending £ Constraint £ Anteriority,
is plotted up in this and all subsequent figures. Fð3;78Þ ¼ 4:68, P , 0:05, e ¼ 0:42; Ending £ Constraint £
Laterality, Fð1;26Þ ¼ 4:26, P , 0:05; Ending £ Constraint £
Laterality £ Hemisphere, Fð1;26Þ ¼ 5:16, P , 0:05). Over-
from anthologies so that their comprehension required all, ERPs to jokes were more negative than non-jokes in a
frame-shifting triggered by the sentence-final noun. Stimuli manner characteristic of the N400: broadly distributed, but
were normed on an off-line cloze task in which a separate with a centro-parietal, slightly right hemisphere focus. The
group of 45 adults completed each joke fragment with the N400 is typically associated with the processing of mean-
first word that came to mind. The cloze probability of the ingful stimuli: the larger its amplitude, the harder the task of
joke endings (percentage of subjects completing the lexical integration [10]. However, these N400 joke effects
sentence with the joke endings) was quite low (3%). Non- were only reliable in high (and not low) constraint sentences
joke endings were constructed by replacing the last word of (post-hoc: Ending, Fð1;27Þ ¼ 14:55, P , 0:001).
jokes with contextually congruent completions that matched Joke effects between 300 and 500 ms demonstrate the
the joke endings for length (6.5 characters; SD ¼ 2), brain’s sensitivity to expectations based on frames, scripts,
frequency (87/million; SD ¼ 146), and cloze probability or schemas retrieved from long-term memory [1,12,13], and
(3%; SD ¼ 2:5). are consistent with other reports that N400 is sensitive to
Experimental sentences in joke and non-joke conditions global aspects of context [15,18]. Although the joke and
were thus identical until the sentence-final word. Non-joke non-joke endings were equally unexpected according to
endings were consistent with the contextually evoked frame, our cloze measure, non-joke endings were designed to be
while joke endings required frame-shifting. Experimental more consistent with the contextually evoked schema, while
sentences were further divided according to sentence joke endings required frame-shifting. Thus, it may be that
constraint, a factor thought to affect the specificity of seman- high constraint sentences allow the reader to commit to a
tic expectations [17], operationalized here as the cloze prob- frame that facilitates the processing of non-joke endings, but
ability of the most popular response for each sentence. A makes joke endings more difficult to integrate.
median split yielded sentence constraints of ,40% (e.g. The ANOVA also revealed reliable effects between 500–
“Statistics indicate that Americans spend 80 million a year 700 and 700–900 ms, but subsequent analysis suggested that
on games of chance, mostly weddings/dice”) and .40% (e.g. the effects varied with participants’ ability to ‘get’ the jokes.
“She read so much about the bad effects of smoking she Good comprehenders (n ¼ 14) averaged 83% correct on the
decided she’d have to give up reading/habit”) for low and true/false questions following the jokes, while poor compre-
high constraint sentences, respectively. henders (n ¼ 14) averaged only 64% correct. On questions
Sentences were presented at a rate of two words/s (200 ms following non-jokes, all but one participant scored 83%
duration; final word ¼ 500 ms) and followed by a true/false correct or better, indicating that the distinction between
question that assessed comprehension, especially whether or good and poor comprehenders reflects difficulty with jokes.
 S. Coulson, M. Kutas / Neuroscience Letters 316 (2001) 71–74 73

ms; joke effects for high constraint stimuli were bilateral

and most evident over the anterior scalp (Fig. 3), while for
low constraint stimuli (Fig. 4), joke effects were larger over
the right and evident only over the posterior scalp (post-hoc:
Constraint £ Ending £ Hemisphere £ Laterality £ Anterior/
Posterior, Fð3;39Þ ¼ 3:4, P , 0:05, e ¼ 0:97). This continued
negativity between 500 and 700 ms may reflect vain
attempts by the participants to search semantic memory
for information that might help them make sense of the
jokes. Between 700 and 900 ms, there were no joke effects.
Among good joke comprehenders, the response to jokes
between 500 and 900 ms post-onset varied across the two
Fig. 2. Voltage maps of mean amplitude of the joke effect (jokes hemispheres: ERPs to jokes were less positive than non-
minus non-jokes) measured 300–500 ms in 14 good (left) and 14 jokes over left lateral sites, an effect we refer to as a
poor (right) joke comprehenders. sustained negativity, and more positive over medial poster-
ior sites (500–700 ms: Ending £ Laterality £ Anterior/
To evaluate ERP differences related to participants’ joke Posterior, Fð3;39Þ ¼ 3:26, P , 0:05, e ¼ 0:83; 700–900 ms:
comprehension, we conducted analyses of mean ERP ampli- Ending £ Laterality £ Anterior/Posterior, Fð3;39Þ ¼ 3:22,
tudes between 300–500, 500–700, and 700–900 ms with the P , 0:05, e ¼ 0:83). Collapsed across ending type, low
between-participants factor, comprehension group, and the constraint stimuli elicited a fronto-central positivity, while
within-participants factors, ending, constraint, and various high constraint stimuli elicited a positivity largest over the
scalp topography factors as above. Between 300 and 500 occipito-parietal scalp. Visual inspection of the data
ms post-word onset, there was no main effect of comprehen- suggests that constraint effects are due primarily to differ-
sion group (F , 1), but there were indications of a more ences in ERPs to joke endings; analysis reveals a marginal
anterior distribution of the joke-related N400 among poor interaction between constraint, ending, and anteriority
than good joke comprehenders (Comprehenders £ (Fð3;39Þ ¼ 3:42, P ¼ 0:08, e ¼ 0:48).
Hemisphere £ Laterality £ Anteriority, Fð3;78Þ ¼ 2:93, P , The goal of the present study was to assess the psycholo-
0.05, e ¼ 0:86; Comprehenders £ Ending £ Hemisphere £ gical reality of frame-shifting, and to differentiate it from
Laterality, Fð1;26Þ ¼ 2:58, P , 0:05; see Fig. 2). Joke effects surprise. Our results, however, do not afford a simple
were also observed between 500–700 and 700–900 ms, mapping from the ERP effects to the surprise and coherence
though their nature differed as a function of comprehension stages of joke comprehension. Both good and poor joke
group (500–700 ms: Fð1;26Þ ¼ 5:82, P , 0:05; 700–900 ms: comprehenders showed greater negativity for joke than
Fð1;26Þ ¼ 9:11, P , 0:01; 500–700 ms: Comprehenders £ non-joke endings, although only in high constraint sentences
Ending £ Constraint £ Hemisphere £ Laterality £ Anterior/ and with differing scalp distributions. In good comprehen-
Posterior, Fð3;78Þ ¼ 3:43, P , 0:05, e ¼ 0:71). ders, joke endings in high constraint sentences elicited a
Among poor comprehenders, jokes elicited less positive posterior positivity, reminiscent of the P3b [5,6], which
(more negative) ERPs than non-jokes between 500 and 700 thus may reflect the violation of frame-level expectations in

Fig. 3. Grand average ERPs to joke (dotted) and non-joke (solid) Fig. 4. Grand average ERPs to joke (dotted) and non-joke (solid)
endings to high constraint sentences in good joke comprehen- endings to low constraint sentences in good (left) and poor
ders (n ¼ 14) and poor joke comprehenders (n ¼ 14). (right) joke comprehenders.
74 S. Coulson, M. Kutas / Neuroscience Letters 316 (2001) 71–74

the jokes (Fig. 3). In contrast, the joke endings in low [5] Coulson, S., King, J.W. and Kutas, M., ERPs and domain
constraint sentences elicited a fronto-central positivity (Fig. specificity: beating a straw horse, Lang. Cogn. Process.,
13 (1998) 653–672.
4) which may be related to the novelty P3, or P3a, an anterior [6] Donchin, E. and Coles, M.G., Is the P300 component a mani-
positivity elicited by stimuli that evoke an orienting reaction, festation of context-updating? Behav. Brain Sci., 11 (1988)
and thought to originate in the superior temporal cortex [7,8]. 357–427.
Jokes also elicited a sustained negativity over anterior left [7] Escera, C., Alho, K., Schroeger, E. and Winkler, I., Involun-
lateral sites for both high and low constraint sentences tary attention and distractibility as evaluated with event-
related brain potentials, Audiol. Neuroontol., 5(3–4) (2000)
which may reflect the frame-shifting needed to re-establish 151–166.
coherence. The wave shape and topography of this effect [8] Knight, R.T., Distributed cortical network for visual atten-
differentiate it from both the posterior positivity elicited by tion, J. Cogn. Neurosci., 9 (1997) 75–91.
high constraint jokes, and the anterior positivity elicited by [9] Kutas, M. and King, J.W., The potentials for basic sentence
low constraint jokes. Rather, it resembles a large, slow- processing: differentiating integrative processes, In T.
Innue and J. McClelland (Eds.), Attention and Performance
rising positive drift at the same left anterior sites, reported XVI: Information Integration in Perception and Communi-
by Kutas and King [9], linked to working memory opera- cation, MIT Press, Cambridge, MA, 1996.
tions involved in the construction of a mental model. The [10] Kutas, M., Federmeier, K.D., Coulson, S., King, J.W. and
sustained effect observed here may reflect negative modula- Muente, T., Language, In J. Cacioppo, L.G. Tassinary,
tion of this slow positive drift due to working memory and G.G. Bernston (Eds.), Handbook of Psychophysiology,
2nd Edition, Cambridge University Press, Cambridge; New
operations involved in joke comprehension. York, 2000.
While the positivities have been argued to reflect a [11] Nahemow, L., Humor as a data base for the study of aging,
surprise component of joke processing, and the sustained In L. Nahemow, K.A. McCluskey-Fawcett, and P.E. McGhee
negativity to index frame-shifting needed to re-establish (Eds.), Humor and Aging, Academic Press, Orlando, FL,
coherence, all three effects occur within the same time 1986, pp. 3–26.
[12] Rumelhart, D.E. and Ortony, A., The representation of
window. Temporal overlap of these joke-related ERP effects knowledge in memory, In R.C. Anderson, R.J. Spiro and
thus make it unlikely that joke processing can be accounted W.E. Montague (Eds.), Schooling and The Acquisition of
for in terms of a simple two-stage model with surprise and Knowledge, Lawrence Erlbaum, Hillsdale, NJ, 1977.
coherence engaged in sequence. [13] Schank, R.C. and Abelson, R.P., Scripts, Plans, Goals, and
Understanding: An Inquiry into Human Knowledge Struc-
tures, Lawrence Erlbaum, Hillsdale, NJ, 1977.
S.C. received support from National Research Service [14] Shammi, P. and Stuss, D.T., Humour appreciation: a role of
Award F32 DC00355. Research supported by the following the right frontal lobe, Brain, 122 (1999) 657–666.
grants to M.K.: HD22614; MH52893; AG08313. [15] St. George, M., Mannes, S. and Hoffman, J.E., Global
semantic expectancy and language comprehension, J.
[1] Barsalou, L.W., Frames, concepts, and conceptual fields, In Cogn. Neurosci., 6 (1994) 70–83.
A. Lehrer and E.F. Kittay (Eds.), Frames, Fields, and [16] Suls, J.M., A two-stage model for the appreciation of jokes
Contrasts: New Essays in Lexical and Semantic Organiza- and cartoons: an information processing analysis, In J.H.
tion, Lawrence Erlbaum, Hillsdale, NJ, 1992. Goldstein and P.E. McGhee (Eds.), The Psychology of
[2] Bihrle, A., Brownell, H. and Gardner, H., Comprehension of Humor: Theoretical Perspectives and Empirical Issues,
humorous and nonhumorous materials by left- and right- Academic Press, New York, 1972, pp. 81–100.
brain damaged patients, Brain Cogn., 5(4) (1986) 399–411. [17] Titone, D., Hemispheric differences in context sensitivity
[3] Brownell, H., Michel, D., Powelson, J. and Gardner, H., during lexical ambiguity resolution, Brain Lang., 65 (1998)
Surprise but not coherence: Sensitivity to verbal human 361–394.
in right-hemisphere patients, Brain Lang., 18 (1983) 20–27. [18] Van Berkum, J.J.A., Hagoort, P. and Brown, C.M., Semantic
[4] Coulson, S., Semantic Leaps: Frame-shifting and Concep- integration in sentences and discourse: evidence from the
tual Blending in Meaning Construction, Cambridge Univer- N400, J. Cogn. Neurosci., 11 (1999) 657–671.
sity Press, Cambridge; New York, 2000.