21.01.2015 Views

Superadditivity in multisensory integration: putting the ... - CI Wiki

Superadditivity in multisensory integration: putting the ... - CI Wiki

Superadditivity in multisensory integration: putting the ... - CI Wiki

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

NEUROREPORT<br />

STANFORD AND STEIN<br />

so<br />

superadditivi<br />

ty might not<br />

be a stable<br />

feature of a<br />

neuron, but<br />

dependent of<br />

<strong>the</strong><br />

properties of<br />

<strong>the</strong> stimuli.<br />

enhancements. Do<strong>in</strong>g so has been particularly valuable, if<br />

not essential, for assess<strong>in</strong>g how normal development<br />

[18,19], altered sensory experience [20], or acute perturbations<br />

of <strong>in</strong>put (e.g. cortical <strong>in</strong>activation) [21] affect <strong>the</strong><br />

<strong>in</strong>cidence of <strong>multisensory</strong> <strong>in</strong>tegration with<strong>in</strong> <strong>the</strong> superior<br />

colliculus, but with <strong>the</strong> untoward effect of creat<strong>in</strong>g a biased<br />

representation of superior colliculus <strong>multisensory</strong> <strong>in</strong>teractions<br />

with<strong>in</strong> <strong>the</strong> literature.<br />

In an effort to more fully characterize <strong>the</strong> nature of<br />

<strong>multisensory</strong> <strong>in</strong>tegration for s<strong>in</strong>gle neurons and populations<br />

of neurons with<strong>in</strong> <strong>the</strong> cat superior colliculus, recent studies<br />

have eschewed <strong>the</strong> near-exclusive use of m<strong>in</strong>imally effective<br />

stimuli <strong>in</strong> favor of parametric manipulations of stimulus<br />

<strong>in</strong>tensity to modulate effectiveness over a broad range. In<br />

one such study, Stanford et al. [22] quantitatively evaluated<br />

<strong>the</strong> operation performed by superior colliculus <strong>multisensory</strong><br />

neurons for comb<strong>in</strong>ations of modality-specific stimuli<br />

cover<strong>in</strong>g a wide range of efficacies. Consistent with <strong>the</strong><br />

many <strong>in</strong>dividual examples <strong>in</strong> <strong>the</strong> literature (e.g. see Figs 1<br />

and 2c) this study verified that superadditivity is <strong>in</strong> fact<br />

commonly observed when very <strong>in</strong>effective modality-specific<br />

stimuli are comb<strong>in</strong>ed. They also, however, found that <strong>the</strong><br />

<strong>in</strong>cidence of superadditive <strong>in</strong>teractions fell precipitously<br />

with <strong>in</strong>creas<strong>in</strong>g efficacy of <strong>the</strong> <strong>in</strong>dividual modality-specific<br />

stimulus components. Indeed, across <strong>the</strong> broader range of<br />

efficacies, <strong>the</strong> majority of <strong>the</strong> <strong>in</strong>teractions <strong>in</strong> <strong>the</strong>ir sample<br />

approximated l<strong>in</strong>ear summation of <strong>the</strong> modality-specific<br />

<strong>in</strong>puts, with superadditive and subadditive <strong>in</strong>teractions<br />

def<strong>in</strong><strong>in</strong>g <strong>the</strong> tails of a normal distribution (Fig. 3).<br />

Analogous results were described by Perrault et al. [23],<br />

who also evaluated <strong>multisensory</strong> <strong>in</strong>teractions aga<strong>in</strong>st a<br />

benchmark of additivity with emphasis on relationships<br />

between <strong>the</strong> <strong>multisensory</strong> operation and <strong>the</strong> dynamic<br />

ranges of <strong>in</strong>dividual superior colliculus neurons. The results<br />

were once aga<strong>in</strong> consistent with extant examples <strong>in</strong> <strong>the</strong><br />

literature, but also established <strong>the</strong> context <strong>in</strong> which such<br />

previous examples should be considered. Perrault et al. [23]<br />

demonstrated that weakly responsive neurons with very<br />

compressed dynamic ranges (i.e. weakly responsive over a<br />

broad range of stimulus <strong>in</strong>tensities) were those most likely<br />

to demonstrate superadditivity exclusively, whereas those<br />

with more expansive and more l<strong>in</strong>ear dynamic ranges<br />

tended to transition from superadditivity to additivity or<br />

from additivity to subadditivity as stimulus <strong>in</strong>tensity (and<br />

efficacy) <strong>in</strong>creased. These recent studies demonstrated that,<br />

<strong>in</strong> <strong>the</strong> superior colliculus at least, superadditivity is but one<br />

facet of <strong>multisensory</strong> <strong>in</strong>tegration, and one that is produced<br />

under a very circumscribed range of circumstances, specifically<br />

when <strong>the</strong> unisensory stimuli to be comb<strong>in</strong>ed are<br />

weakly effective. As one might have expected, <strong>the</strong> results of<br />

<strong>the</strong>se studies illustrate that <strong>the</strong> pr<strong>in</strong>ciple of ‘<strong>in</strong>verse<br />

effectiveness’ can be extended beyond <strong>multisensory</strong> enhancement<br />

to <strong>in</strong>clude <strong>the</strong> form of <strong>the</strong> <strong>multisensory</strong><br />

computation.<br />

These larger surveys of superior colliculus <strong>in</strong>tegration are<br />

particularly germane to consideration of an earlier study<br />

by Popul<strong>in</strong> and Y<strong>in</strong> [24], which stands out as <strong>the</strong> only<br />

proponent of a contrarian and ra<strong>the</strong>r provocative conjecture<br />

that superadditivity is an artifact of <strong>the</strong> anes<strong>the</strong>tic agent<br />

used <strong>in</strong> most previous studies. This conclusion was based<br />

on <strong>the</strong> fact that <strong>in</strong> <strong>the</strong>ir study of <strong>multisensory</strong> <strong>in</strong>tegration <strong>in</strong><br />

<strong>the</strong> superior colliculus of alert cats, Popul<strong>in</strong> and Y<strong>in</strong> found<br />

little evidence of superadditivity, <strong>in</strong>stead report<strong>in</strong>g that<br />

most <strong>multisensory</strong> <strong>in</strong>teractions were ei<strong>the</strong>r additive (l<strong>in</strong>ear)<br />

Response mode proportion<br />

100<br />

80<br />

60<br />

40<br />

20<br />

0<br />

0−2 2−4 4−6 6−8<br />

Superadditive<br />

Additive<br />

Subadditive<br />

8−10 10−12 12−14<br />

Unisensory response sum (impulses/trial)<br />

Fig. 3 The <strong>in</strong>cidence of superadditivity decl<strong>in</strong>es rapidly with <strong>in</strong>creas<strong>in</strong>g<br />

stimulus strength. Shown here, <strong>in</strong> a ¢gure adapted from Figure 6a of Stanford<br />

et al. [22], is an illustration of how <strong>the</strong> neural <strong>in</strong>tegration of crossmodal<br />

stimulus pairs depends on <strong>the</strong> strength of <strong>the</strong> modality-speci¢c<br />

component stimuli. In this experiment, s<strong>in</strong>gle neuron activity from <strong>the</strong><br />

cat superior colliculus was recorded <strong>in</strong> response to visual, auditory, and<br />

visual-auditory stimuli across a wide range of stimulus <strong>in</strong>tensities. In each<br />

case, <strong>the</strong> magnitude of <strong>the</strong> response to <strong>the</strong> cross-modal stimulus was<br />

evaluated with respect to a benchmark of simple summation of <strong>the</strong> responses<br />

to <strong>the</strong> modality-speci¢c component stimuli (see Stanford et al.<br />

[22] for details).The plot illustrates <strong>the</strong> relative likelihood of <strong>multisensory</strong><br />

responses that exceeded summation (superadditive¼¢lled circles), failed<br />

to achieve summation (subadditive¼open squares), or were consistent<br />

with summation (additive¼open circles). Note that <strong>the</strong> likelihood of observ<strong>in</strong>g<br />

superadditivity fell dramatically as <strong>the</strong> e⁄cacy of <strong>the</strong> component<br />

stimuli <strong>in</strong>creased. Thus, superadditivity predom<strong>in</strong>ated only for very<br />

weakly e¡ective auditory and visual stimulus components, stimuli for<br />

which <strong>the</strong> predicted sum of <strong>the</strong> modality-speci¢c responses failed to<br />

exceed 2^ 4 impulses/trial.<br />

or subadditive (subl<strong>in</strong>ear). Although <strong>the</strong> Popul<strong>in</strong> and Y<strong>in</strong><br />

f<strong>in</strong>d<strong>in</strong>gs appeared to be an extreme departure from earlier<br />

literature, <strong>in</strong>clud<strong>in</strong>g those from o<strong>the</strong>r studies <strong>in</strong> awake<br />

animals (e.g. see Ref. [25]), <strong>the</strong>ir data overlap greatly with<br />

<strong>the</strong> more recent studies show<strong>in</strong>g that, <strong>in</strong> anes<strong>the</strong>tized cats,<br />

additivity predom<strong>in</strong>ates for comb<strong>in</strong>ations of all but <strong>the</strong><br />

weakest modality-specific stimuli [22,23]. Popul<strong>in</strong> and Y<strong>in</strong><br />

do not relate <strong>in</strong>tegration mode to stimulus efficacy for <strong>the</strong>ir<br />

sample; however, <strong>the</strong> specific examples provided suggest<br />

moderate efficacy and, at first glance, seem <strong>in</strong> l<strong>in</strong>e with <strong>the</strong><br />

most recent results from anes<strong>the</strong>tized cats. This, along with<br />

both earlier [25] and more recent reports of superadditive<br />

<strong>multisensory</strong> <strong>in</strong>teractions <strong>in</strong> s<strong>in</strong>gle neurons <strong>in</strong> a variety of<br />

structures <strong>in</strong> awake, behav<strong>in</strong>g animals (monkey cortex: [26];<br />

rat thalamus: [27]), coupled with <strong>the</strong> data from fMRI and<br />

ERP studies (see above) strongly suggest that a difference <strong>in</strong><br />

sampl<strong>in</strong>g and/or stimulus efficacy is <strong>the</strong> more parsimonious<br />

explanation of <strong>the</strong> extreme paucity of superadditive<br />

cases <strong>in</strong> <strong>the</strong>ir sample.<br />

Implications for behavior<br />

From a functional perspective, <strong>the</strong> issue of <strong>in</strong>tegrative<br />

mechanism (i.e. l<strong>in</strong>ear or nonl<strong>in</strong>ear) is relevant only to <strong>the</strong><br />

extent that it dictates <strong>the</strong> magnitude (and/or tim<strong>in</strong>g) of <strong>the</strong><br />

postsynaptic response. Neurons <strong>in</strong> <strong>the</strong> superior colliculus<br />

represent salient visual, auditory, and tactile stimuli and<br />

contribute to <strong>the</strong> formation of motor commands to orient<br />

even if a<br />

certa<strong>in</strong> % of<br />

trials with<br />

superadd,<br />

that would<br />

manifest as<br />

<strong>in</strong>creased<br />

activity <strong>in</strong><br />

study..<br />

79 0 Vol 18 No 8 28 May 2007<br />

Copyright © Lipp<strong>in</strong>cott Williams & Wilk<strong>in</strong>s. Unauthorized reproduction of this article is prohibited.

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!