{"id":231,"date":"2020-11-24T11:21:59","date_gmt":"2020-11-24T11:21:59","guid":{"rendered":"http:\/\/www2.mrc-lmb.cam.ac.uk\/groups\/tripodi\/?p=231"},"modified":"2020-11-24T11:23:18","modified_gmt":"2020-11-24T11:23:18","slug":"chomsky-and-the-platypus","status":"publish","type":"post","link":"https:\/\/www2.mrc-lmb.cam.ac.uk\/groups\/tripodi\/2020\/11\/24\/chomsky-and-the-platypus\/","title":{"rendered":"Chomsky and the platypus"},"content":{"rendered":"\n

Sometime you read a paper and you keep thinking about it. That\u2019s what happened to me with the latest work from Xin Jin\u2019s lab<\/a> (yes, I am in awe of this work) in which they show that there exists a hierarchical organization of learned action sequences<\/a> .<\/p>\n\n\n\n

First of all, what\u2019s an action sequence and what\u2019s the problem? Let\u2019s assume I need to do ABCD to obtain a given reward (i.e. [toss the ball (A)], [bring the racket shoulder height (B)], [extend the arm(C)], [hit the ball(D)] -> become a millionaire (reward)). How do I create a mental representation of this sequence so to be able to perform it? I might generate a representation of the linear sequence of action-tokens <\/em>(\u201cA\u201d->\u201dB\u2019\u201d->\u201dC\u201d->\u201dD\u2019\u201d) that leads to the reward (where a token <\/em>would be a minimal action element that constitutes a finite action, such as in our example [toss the ball] ). Alternatively, my mental representation could entail a clustering of action-tokens <\/em>into meaningful \u201caction-set<\/em>\u201d, resulting from the Merge of multiple tokens<\/em>, i.e. \u201c{A,B}\u201d such as [toss the ball and bring the racket shoulder height] -> \u201c{C,D}\u201d [extend the arm and hit the ball]. In other words, in my mental construction, either single action-tokens <\/em>bear meaning or it is only the action-set <\/em>such as \u201c{A,B}\u201d that does (we\u2019ll discuss later on what meaning in the action realm might mean, pun not intended).<\/p>\n\n\n\n

Xin Jin and collaborators, with a series of elegant experiments show that the latter is true, at least from a basal-ganglia-centric point of view. Very briefly (as for once I don\u2019t intend to discuss the experimental part in great detail so to have time to venture a bit outside the main scope of the work), what they did was to train mice to perform a defined action sequence<\/a>, consisting of an ordered series of lever presses, in order to obtain a reward (the accent here is on ordered). Such an action sequence could be composed, for instance, of two L<\/strong>eft lever presses followed by two R<\/strong>ight ones (LLRR, see figure below). They then manipulated the activity of basal ganglia neurons of either the direct or indirect pathway and assesed how this impacted on the ongoing action sequence. I will argue that manipulating these pathways and test what happens to the individual action-tokens <\/em>of the action-sequence is a way to ask the mouse to define what are the foundamental elements of such a series (i.e are they the action-tokens <\/em>or the action-set<\/em>?, see also the concept of “Chunking of action repertoires” in Graybiel, 1998<\/a>), as only such elements will be affected by the manipulation (if you don\u2019t agree with this, you\u2019ll disagree with all that follows).<\/p>\n\n\n\n

\"\"<\/figure><\/div>\n\n\n\n

What does this imply in broader terms? [Disclaimer, what follows is on me as the authors, with good reasons, don\u2019t venture in the nonsense I am about to embark on].<\/p>\n\n\n\n

Albeit not explicitly claimed or discussed by the authors, reading this work it\u2019s hard to resist to the allure of the idea of the existence of a syntactic organization of actions. There is no doubt that the hierarchical organization of learned action sequence uncovered by Xin Jin and collaborators bears striking resemblance with Chomsky\u2019s hierarchical structure of language (Chomsky, 1957<\/a>, see figure below) [hence half of the title of this post\u2026 the other half is there only because I love Umberto Eco’s essay on Semiotics<\/a>].<\/p>\n\n\n\n

\"\"<\/figure><\/div>\n\n\n\n

Now, I know that I am not being very original as such an analogy has already been suggested multiple times in the past (Pulvermuller and Fadiga 2010<\/a>; or think of the \u201chigh order vocabulary\u201d in Rizzolatti et al., 1988<\/a>), but, as correctly pointed out previously (Moro, 2014<\/a>), \u201cis this analogy between language and actions true, or, at least, useful?\u201d.<\/p>\n\n\n\n

Two main criticisms are usually associated with the analogy between language and actions (indeed more than two\u2026 but for brevity let\u2019s focus on the key ones): 1<\/strong>) the lack of action equivalents of \u2018functional words\u2019 such as those containing logical instructions (i.e. \u201cif\u201d, \u201cthen\u201d, \u201cnot\u201d); 2<\/strong>) The existence (or lack thereof) of principles governing parsing (the assignment of a syntactic structure to recursively merged lexical items) of an action sequence (action-phrase<\/em>).<\/p>\n\n\n\n

With respect to the first, it has always been my impression that this, really, is a non-problem. In the action realm, logical instructions are implicit in the motor sequence and reflect the learning paradigm related to the specific sequence. For example, in the task devised by Xin Jin and colleagues, mice have been trained to do 2x L<\/strong>eft ever presses (\u201cLL\u201d) followed by 2x R<\/strong>ight lever presses (\u201cRR\u201d) in order to obtain a reward. Now, from perspective of the animal, such a sequence can be algorithmically implemented as: \u201cif\u201d I [press the left lever twice] \u201cthen\u201d [[press right lever twice] \u201cand\u201d [\u201cdo not\u201d press left again]] I\u2019ll be rewarded. I guess you\u2019ll agree that, in such an action-lexicon<\/em>, functional words (\u201cif\u201d, \u201cthen\u201d, \u201cnot\u201d) are implicit in the learned motor sequence. Indeed, one could even argue that the ordered sequence of the actions taken (and of all of those not taken) is the physical implementation of such logical operators \u2013 and that\u2019s it for point one.<\/p>\n\n\n\n

With respect to the second criticism the situation is a bit more complex, but this is where in a final, brilliant and somehow understated experiment, Xin Jin\u2019s work shines. In linguistics, the process of Merge is one in which lexical items are combined, potentially ad infinitum. However, possibly due to limits to memory capacity or time pressure on decoding, strategies to simplify the identification of underlying syntactic structure seem necessary. One of such strategies consists in parsing by \u201cminimal attachment\u201d principles (the parser builds the simplest syntactic structure possible, the one with few nodes). \u201cMarco spoke to Letizia and Laura replied that\u2026\u201d, in this example of Merge the reader has to understand who is the subject of spoke and replied. It\u2019s clear that the process of parsing is intimately linked to\/derived by the nature of language itself and that by parsing a sentence we reveal something about the logic of how it is generated. By reflection, one could say that if we understand how action-phrases <\/em>are parsed we\u2019ll understand something about how they are generated.<\/p>\n\n\n\n

The Frazier\u2019s serial modular model of parsing (the so called garden path model) postulates that one first defines the structure of the sentence by principles of minimal attachment and late closure (the principle by which incoming words tend to be associated with the phrase currently being processed), and that only at a later point one refines it according to the semantic meaning of its components (I ask what it actually means).<\/p>\n\n\n\n

At the other end of the spectrum, there is the idea that, in order to correctly parse a sentence, a contextual and semantic understanding must take place first.<\/p>\n\n\n\n

So, do Xin Jin and colleague tell us something about whether modular or semantic parsing occur at all in the action realm? I\u2019d say yes, to some extent at least, but this is where I am stretching my argument very thin. In the action-phrase <\/em>LLRR, principle of Locality might lead to clustering \u201cLL\u201d together and hence, \u201cRR\u201d. But what if I add an extra \u201cL\u201d and an extra \u201cR\u201d; how do I parse this new action-phrase <\/em>then? This is what Xin Jin and collaborators did by training mice on an [LLLRRR] task. It turns out that mice now see \u201cLLL\u201d and \u201cRRR\u201d as action-sets <\/em>(always in a basal-ganglia-centric point of view), as opposed to \u201cLL\u201d \u201cLR\u201d \u201cRR\u201d as it might have emerged from Locality principles alone. In other words, it seems to me, that basal ganglia do care about the semantic of the action and that such semantic is assigned, in the action realm, on the basis of learned outcomes; hereby, in the latter contingency (LLLRRR), \u201cLL\u201d ceases having a semantic meaning as no learned outcome is linked to it, while \u201cLLL\u201d gains a semantic meaning by means of the reward linked to its implementation.<\/p>\n\n\n\n

\"\"<\/figure><\/div>\n\n\n\n

Now, I need to stress that experiments were not designed with a linguistic angle in mind (I guess) and hence they are not immediately suited for making these points, yet I still think that it\u2019s interesting (at least fun) to think about it in these terms.<\/p>\n\n\n\n

Hence, with respect to this action-language analogy, three things emerge from this study, the first of which very clearly, while the other two less so: firstly, action sequences (action-phrases<\/em>) are hierarchically organised (which is indeed the main claim of the authors); secondly, parsing action-phrases <\/em>(or generating\/crystallising them) requires some element of semantic understanding; finally, semantic understanding in the context of actions would be linked to some reward probability function (if a given action-set <\/em>impacts on the statistics of the world, that would give the set <\/em>a meaning). Clearly the last two points are far from being proven and new experiments could be designed, starting from the current experimental design, in order to address them more directly.<\/p>\n\n\n\n

In summary, it\u2019s probably early to say whether the analogy between language and action is true, however, going back to Moro\u2019s question, I believe that it might be indeed a very useful one.<\/p>\n","protected":false},"excerpt":{"rendered":"

Sometime you read a paper and you keep thinking about it. That\u2019s what happened to me with the latest work from Xin Jin\u2019s lab (yes, I am in awe of this work) in which they show that there exists a hierarchical organization of learned action sequences . First of all, what\u2019s an action sequence and what\u2019s the problem? Let\u2019s […]<\/p>\n","protected":false},"author":11,"featured_media":233,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_genesis_hide_title":false,"_genesis_hide_breadcrumbs":false,"_genesis_hide_singular_image":false,"_genesis_hide_footer_widgets":false,"_genesis_custom_body_class":"","_genesis_custom_post_class":"","_genesis_layout":"","footnotes":""},"categories":[8],"tags":[],"class_list":{"0":"post-231","1":"post","2":"type-post","3":"status-publish","4":"format-standard","5":"has-post-thumbnail","7":"category-blog","8":"entry"},"featured_image_src":"https:\/\/www2.mrc-lmb.cam.ac.uk\/groups\/tripodi\/wp-content\/uploads\/sites\/15\/2020\/11\/innercomp_txtMedialwgimgimage.jpg","featured_image_src_square":"https:\/\/www2.mrc-lmb.cam.ac.uk\/groups\/tripodi\/wp-content\/uploads\/sites\/15\/2020\/11\/innercomp_txtMedialwgimgimage.jpg","author_info":{"display_name":"matt","author_link":"https:\/\/www2.mrc-lmb.cam.ac.uk\/groups\/tripodi\/author\/matt\/"},"_links":{"self":[{"href":"https:\/\/www2.mrc-lmb.cam.ac.uk\/groups\/tripodi\/wp-json\/wp\/v2\/posts\/231","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www2.mrc-lmb.cam.ac.uk\/groups\/tripodi\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www2.mrc-lmb.cam.ac.uk\/groups\/tripodi\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www2.mrc-lmb.cam.ac.uk\/groups\/tripodi\/wp-json\/wp\/v2\/users\/11"}],"replies":[{"embeddable":true,"href":"https:\/\/www2.mrc-lmb.cam.ac.uk\/groups\/tripodi\/wp-json\/wp\/v2\/comments?post=231"}],"version-history":[{"count":2,"href":"https:\/\/www2.mrc-lmb.cam.ac.uk\/groups\/tripodi\/wp-json\/wp\/v2\/posts\/231\/revisions"}],"predecessor-version":[{"id":236,"href":"https:\/\/www2.mrc-lmb.cam.ac.uk\/groups\/tripodi\/wp-json\/wp\/v2\/posts\/231\/revisions\/236"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www2.mrc-lmb.cam.ac.uk\/groups\/tripodi\/wp-json\/wp\/v2\/media\/233"}],"wp:attachment":[{"href":"https:\/\/www2.mrc-lmb.cam.ac.uk\/groups\/tripodi\/wp-json\/wp\/v2\/media?parent=231"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www2.mrc-lmb.cam.ac.uk\/groups\/tripodi\/wp-json\/wp\/v2\/categories?post=231"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www2.mrc-lmb.cam.ac.uk\/groups\/tripodi\/wp-json\/wp\/v2\/tags?post=231"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}