Talk:Ocular dominance column
The introductory paragraph, though fine in itself, offers a longer lead-in to the topic than I think is required for this kind of encyclopedia article. What would fit the 'wiki style' better would be to define the topic title 'ocular dominance' in the first sentence or two, together with an explanation of the Hubel and Wiesel 7 point scale (which is basic and still widely used). Ocular dominance columns (which are not quite the same thing) could then be defined and some comments of a general nature about them could be given – e.g. how they were first discovered and their occurrence and varied morphology in monkeys and cats (as shown in Fig. 1).
Section 1 ("Development') gives the impression that odcs only occur within layer IV which is not the case. They are less well defined in the upper and lower layers – perhaps this could be explained.
The end of this section does not clearly distinguish LGN layer segregation from cortical segregation. Also, since a lot of work is now done on ferrets, I think it would be appropriate to include time lines for LGN and cortical segregation for ferrets as well as cats. Figure 1 of a review by Sengpiel and Kind (Current Biology, 2002) offers a very useful summary of the relevant time lines and might perhaps be included as a figure.
Section 1.1 regarding functionality might cite (or mention) Horton's more recent results on ocular dominance variability in squirrel monkeys (Horton and Adams, 2005, Phil. Trans. R. Soc. Lond. B, 360, 837).
Section 2 ("Critical Period") The speculations about the relation between brain weight and life span and the critical period are very interesting but I think they belong at the end of the section rather than the beginning as they are of a more speculative nature. I would begin with the basics and progress to more detail – my own preference would be to be historical and give the observations by Wiesel and Hubel that led to the concept of the critical period and then flesh that out with reference to critical periods for other systems and modalities (e.g. direction preference and perhaps imprinting in birds?). It doesn't have to be historical but I think this first paragraph should give the actual timings of the critical periods for monocular deprivation in cats, monkeys and humans – it doesn't at present. There is a figure somewhere out there showing the effects of varying periods of deprivation at various time periods in different species – maybe from the work of Daphne Maurer, or perhaps Don Mitchell or Marty Steinbach (I am sorry I can't find a specific source right now) – which might be worth including.
Early experiments on segregation and the effects of dark rearing used the trans-neuronal autoradiography technique which is now suspected of seriously underestimating the amount of segregation present, especially in young animals. The extent of this, and the reasons for it are still not understood but probably any experiment using trans-neuronal autoradiography to demonstrate a lack of segregation is suspect. The more important conclusion is that columns can form before birth or eye-opening, which clearly limits the range of possible mechanisms to those not involving visually driven patterns of retinal activity.
Experimental evidence that blocking retinal waves blocks LGN segregation could be cited (e.g. Torborg & Feller, 2005; Prog. Neurobiol., 76, 213).
Section 3 – a bit beyond my expertise to comment on.
Section 4.1. There is a paper by Goodhill on the different kinds of normalisation which could be cited.
Section 5 – this equation predicts LTP when both terms are –ve. Miller (1996) says this can just be ignored but it perhaps needs to be spelt out, or formalised.
I have not checked the details of the equations and descriptions of the models of Harris et al. and Elliott & Shadbolt. Scholarpedia has a strong theoretical emphasis so this length of coverage of individual models may be appropriate. However my own bias would be to provide more coverage of the experimental side of the topic and describe the various models more briefly – I think it is fair to say that none of them can be regarded as more than possible speculations about the actual mechanisms at the moment, so arguably the empirical coverage will be more useful in a short space.
I have edited a few typos directly.
We thank the reviewer for his suggestions. We have modified the Scholarpedia entry according to these suggestions and we provide a point-by-point reference to those modifications below.
- We have modified the introductory paragraph in order to make it shorter. Now, it directly leads into the main topic as suggested by the reviewer.
- We now emphasize that ocular dominance columns are also apparent outside layer 4 (Section 1, paragraph 1).
- We have also made a clear distinction between LGN segregation and cortical segregation. We have followed the reviewer’s suggestion and added Figure 1 of a review by Sengpiel and Kind (Current Biology, 2002) (see Section 1, figure 4).
- We have included a citation to Horton and Adams, 2005 (Section 1.1).
- We could not find the figure suggested by the reviewer (showing the effects of varying periods of deprivation at various time periods in different species). However, we have modified section 2 following the reviewer suggestions and the discussion emphasizes differences across species.
- We now mention Goodhill’s review (Section 4.1).
- In section 5 we have explicitly mentioned the particular case in which both factors in the right side of the equation are less that zero. We direct the reader to Miller’s work for a full discussion.
Regarding the models, we consider that this is an important component of the Scholarpedia entry. The quantitative models help summarize experimental observations and organize them into a rigorous quantitative framework that allows explaining existing data and make novel predictions for future research. We therefore think that this will be valuable to readers and we decided to leave the discussion of the different computational models.
REVIEWER 1 COMMENTS, ROUND 2 The article is improved now, and the majority of my points have been addressed. However I still feel a need to distinguish between 'ocular dominance' and 'ocular dominance columns' - the two are not the same (i.e. discuss the absence of odcs in rodents and other animals, but note that cells in rats and mice have individual ocular dominance values, probably randomly varying from cell to cell). Also the 7 point scale could be explained in still more detail (perhaps with a figure) - I can imagine someone consulting this article to find out how it works and coming away disappointed.
The spacing for human odcs should probably be given as 1.7 mm and not 1.4 mm (see Adams and Horton, 2006, the most recent and best data on human odcs)
There are a few minor typos which could likely be found by a spelling checker.
I can add that I am in general agreement with the various points raised by Ken Miller below. Sorry I forgot about Miller and MacKay (1994), no reason not to cite it in the context. One caveat though (on the side of the authors) is that an encyclopedia article has to be selective and succinct: it should concentrate on well established findings (and models) as much as possible, and only briefly point out major areas of current ignorance and/or disagreement. It cannot be totally up to date on all details and points of qualification - that is the function of review articles. I.e. an encyclopedia article (as I understand the mandate) should aim to stand the test of time, whereas a review article might need major updating a few years later.
So areas of controversy and uncertainty (like the Weliky-Crowley one, or the issues surrounding the use of trans-neuronal autoradiography) might be best summed up briefly, with a minimal number of references.
REVIEWER 2 (Ken Miller):
Overall I think this article has two major problems. First, it is not very up-to-date on the experiments. Second, the presentation of theory seems to focus on exhaustively listing the equations of the models without telling us much about what was actually learned. I'll try to provide below the material I think is needed to improve this.
With respect to spacing of OD columns, see recent work greatly improving the quantitative analysis and showing greater variability than previously realized in various species:
Kaschube M, Wolf F, Puhlmann M, Rathjen S, Schmidt KF, Geisel T, Löwel S. Eur J Neurosci. 2003 Dec;18(12):3251-66. The pattern of ocular dominance columns in cat primary visual cortex: intra- and interindividual variability of column spacing and its dependence on genetic background.
Rathjen S, Schmidt KE, Löwel S. Exp Brain Res. 2002 Jul;145(2):158-65. Two-dimensional analysis of the spacing of ocular dominance columns in normally raised and strabismic kittens.
Horton JC, Hocking DR. J Neurosci. 1996 Nov 15;16(22):7228-39 Intrinsic variability of ocular dominance column periodicity in normal macaque monkeys.
Adams DL, Sincich LC, Horton JC. J Neurosci. 2007 Sep 26;27(39):10391-403 Complete pattern of ocular dominance columns in human primary visual cortex.
Also you cite Horton et al. 1990 for humans but (a) it's not in your references and (b) that reference speaks of columns ranging from 1 to 2 mm, whereas you say they are 1.4 mm without mentioning the variability.
It's also important to note that not all species have ocular dominance columns. Rodents do not seem to have ocular dominance columns or any other columnar organization in V1 except retinotopy -- see the Gordon and Stryker paper (which you cite but do not have in your list of references) for mouse ocular dominance; the same is true for rats, but you'll have to find a reference, see recent papers from Clay Reid (R.C. Reid) using calcium imaging for lack of orientation columns in mice; squirrels, which have a good visual system, don't have orientation columns despite having plenty of orientation-tuned cells (van Hooser et al, J Neurosci 25:19-28 (2005)), but ocular dominance hasn't been documented in squirrels to my knowledge. See also papers by Jonathan Horton on lack of ocular dominance columns in many squirrel monkeys -- e.g.
Adams DL, Horton JC. Monocular cells without ocular dominance columns. J Neurophysiol. 2006 Nov;96(5):2253-64
Adams DL, Horton JC. Capricious expression of cortical columns in the primate brain. Nat Neurosci. 2003 Feb;6(2):113-4.
DEVELOPMENT OF OCULAR DOMINANCE COLUMNS
This is ignorant of a lot of the recent literature. This is a good recent review that I suggest the authors read:
Huberman AD. Mechanisms of eye-specific visual circuit development. Curr Opin Neurobiol. 2007 Feb;17(1):73-80
Specifically, with respect to ocular dominance columns: J Crowley and LC Katz in two very prominent and influential papers in 1999 and 2000 challenged the idea that ocular dominance columns segregate from initially overlapping projections, arguing instead that the two eyes' inputs grow directly in to segregated patches and by implication that the locations of the two eyes' patches are genetically specified:
Crowley JC, Katz LC. Early development of ocular dominance columns. Science. 2000 Nov 17;290(5495):1321-4
Crowley JC, Katz LC. Development of ocular dominance columns in the absence of retinal input. Nat Neurosci. 1999 Dec;2(12):1125-30
Personally, I think they may well be wrong. They showed that projections from LGN into V1 were patchy from a very early age (I think they looked as early as P9 in a ferret), but I don't believe they really showed that these patches corresponded to ocular dominance, nor did they show that the patches don't develop from an initially more uniform innervation (they looked about a week after LGN afferents arrive in layer 4, leaving plenty of time for activity-dependent processes). Note that connectivity that leads to periodic activity patterns in cortex will lead, via Hebbian mechanisms, to correlated inputs making periodic, patchy projections, so the patches they see could just be local retinotopic regions of LGN serving both eyes making periodic, patchy connections. One key test would be to inject the monocular region of LGN and see if it also makes patchy connections, or if instead it makes continuous connections. I think Huberman also discusses issues with interpretation of these results, but there is little in the literature questioning the Crowley/Katz interpretation or raising the concerns I've just raised. At any rate, right or wrong, they have been very influential and deserve discussion. (Note, Sengpiel's figure (your Figure 4) says "OD patches after LGN injection" for this time for ferrets. This is based on believing that the patches Crowley and Katz saw really were OD patches.)
It is also worth being aware of results from Mike Weliky showing that V1 indeed does have periodic spontaneous activity patterns early in development (P22-P28 in ferrets), well before what had been regarded as the time ocular dominance develops, though they didn't look as early as Katz and Crowley saw patches:
Chiu C, Weliky M. Neuron. 2002 Sep 12;35(6):1123-34. Relationship of correlated spontaneous activity to functional ocular dominance columns in the developing visual cortex.
Chiu C, Weliky M. J Neurosci. 2001 Nov 15;21(22):8906-14 Spontaneous activity in developing ferret visual cortex in vivo.
Note that the later paper saw that physiologically, ocular dominance was already segregating in accord with the periodic activity patterns at this time. That of course does not prove that there is anatomical segregation at this time.
Here is another paper from Weliky to be aware of, showing that, at this same early time, LGN spontaneous activity shows positive correlations between the two eyes, induced by feedback from V1 to LGN (so it's not necessarily the case that spontaneous activities of the two eyes' inputs to V1 are independent, though this might vary with developmental time):
Weliky M, Katz LC. Correlational structure of spontaneous neuronal activity in the developing lateral geniculate nucleus in vivo. Science. 1999 Jul 23;285(5427):599-604
Around the same time as Crowley and Katz' work, Stryker's lab revisited the early development of ocular dominance in cat V1. They found that as early as P14 but not as early as P7, they could see early ocular dominance segregation: the contralateral eye dominated and projected inputs everywhere, while the ipsilateral eye projected to periodic islands in the sea of contralateral inputs. At about the same time as the onset of the critical period for monocular deprivation (MD), the two eyes' projections roughly equalize into the mature pattern of ocular dominance columns, but only if there is visual experience; if the animals do not have visual experience, the OD pattern remains in its immature, ipsi-islands-in-a-contra-sea state (and the critical period for MD also does not open, as many have shown):
Crair MC, Gillespie DC, Stryker MP. The role of visual experience in the development of columns in cat visual cortex. Science. 1998 Jan 23;279(5350):566-70
Crair MC, Horton JC, Antonini A, Stryker MP. Emergence of ocular dominance columns in cat visual cortex by 2 weeks of age. J Comp Neurol. 2001 Feb 5;430(2):235-49
As discussed in the Discussion in the 2001 article, individual geniculocortical axonal arbors at the early ages are quite sparse, and only later branch and widen, unlike earlier pictures that thought that initially broad, continuous arbors then became patchy to yield ocular dominance segregation. Also, I think Figure 10 of the 2001 article probably does a better job than the Sengpiel cat timeline figure. (Another nice timeline figure is Fig. 7 of Issa NP, Trachtenberg JT, Chapman B, Zahs KR, Stryker MP, The critical period for ocular dominance plasticity in the Ferret's visual cortex, J Neurosci. 1999 Aug 15;19(16):6965-78, but unfortunately it is not up to date with the Crair and Stryker or Crowley and Katz work).
FUNCTIONALITY OF OCULAR DOMINANCE COLUMNS
We know that mice and rats (and probably squirrels) have retinotopy and binocular regions in V1 but don't have ocular dominance segregation. So it's wrong to suggest that two retinas plus retinotopy implies OD segregation. Clearly one thing that is needed is either genetic specification or, if it is self-organized, then neurons must influence their neighbors' development, presumably through local excitatory and inhibitory connections, in order that localized patches of neurons develop common properties. At least rough retinotopy is clearly specified genetically, but if rodents don't have sufficient inter-neuron interactions during development, then any self-organized properties will come out salt-and-pepper, as orientation and ocular dominance selectivity apparently do.
CRITICAL PERIOD AND THE ROLE OF ELECTRICAL ACTIVITY
As the work of Crair and Stryker made clear and the Crowley and Katz work also strongly argued, the onset of ocular dominance segregation occurs significantly before the opening of the critical period for MD. The original picture, that ocular dominance segregation begins at the onset of the critical picture and is part of the same process, is incorrect. Much of this section discusses the onset of ocular dominance segregation rather than the critical period. That material would be much better placed in the section on "Development of ocular dominance columnns".
The results that dark rearing leads to reduced or abnormal segregation in cats I think is now better understood from the Crair and Stryker results: without visual experience, the OD segregation remains in its initial, immature state. Monkeys are born with a mature pattern of OD columns, as first shown by the LeVay et al 1980 work you reference but perhaps shown most definitively recently by Horton and Hocking 1996 (you now reference this paper but in another context).
Strangely, you cite Gordon and Stryker (1996) and Horton and Hocking (1996) as calling into question results that failed to find segregation using trans-neuronal autoradiography. Horton and Hocking actually used trans-neuronal autoradiography in their paper, and Gordon and Stryker is as far as I can tell irrelevant to this issue. It was the Crowley and Katz papers that first strongly called such results into question, as they were arguing that they were seeing segregation much earlier than it had been seen in the classic autoradiography papers. They argued that spillover, particularly in young animals, might have led those autoradiography papers to fail to see the segregation at young ages that they were now seeing. Spillover refers to label injected into one eye spilling over in LGN to and labelling the other eye's layer, leading spuriously to uniform labelling in V1 even though a single eye's terminations in V1 would show segregation. The Crair and colleagues 2001 paper also specifically addressed this issue: they argued that careful transneuronal labelling gives accurate results that match what's found with other techniques, although earlier transneuronal studies had not been as sensitive, see first paragraph of their Discussion. It's worth noting that the first developmental study of OD columns using transneuronal autoradiography had an extensive discussion of the spillover problem and attempted to quantitatively account for it (LeVay S, Stryker MP, Shatz CJ. Ocular dominance columns and their development in layer IV of the cat's visual cortex: a quantitative study. J Comp Neurol. 1978 May 1;179(1):223-44) -- they might have got the answer wrong, but they were well aware of and struggled to deal with the problem.
You mention the Stryker et al. 1986 paper. There is a remaining concern about it, raised by the Crowley and Katz picture, that perhaps there was segregation before they injected the TTX and the TTX leads to a loss of segregation, in which case the TTX results would not tell us whether the *initial* segregation was activity-dependent. I don't believe this, but it can't be ruled out. Stryker started injecting as early as P12, the Crair et al. results show that some segregation is apparent a couple of days later than this but not 5 days before.
You also say that OD columns don't form when retinal waves are disrupted, but don't give a reference. Perhaps you are thinking of this paper:
Huberman AD, Speer CM, Chapman B. Spontaneous retinal activity mediates development of ocular dominance columns and binocular receptive fields in v1. Neuron. 2006 Oct 19;52(2):247-54.
Note that *all* retinal activity was blocked, this wasn't just a disruption of the waves.
You discuss retinal waves in several sentences and then only at the end of the same paragraph do you explain that there are waves of activity that propagate across the early retina. Obviously the description should come before other discussion. Note that these waves disappear well before the initial segregation of OD columns as measured by Crair et al. in cats, if not as measured by Crowley and Katz in ferret. But Huberman et al 2006 found that blocking retinal activity at these very early ages in ferret when waves still exist disrupts the ultimate formation of OD columns.
It's not necessarily true that the waves reaching V1 are likely to be asynchronous in the two eyes, see previous discussion above of Weliky's results on early LGN correlations.
Note that the critical period is the time at which *ocular dominance* becomes susceptable to shifting by visual experience, but before the critical period other aspects of visual cortical development can be changed by altering visual experience and in particular by monocular deprivation:
Smith SL, Trachtenberg JT. Experience-dependent binocular competition in the visual cortex begins at eye opening. Nat Neurosci. 2007 Mar;10(3):370-5
MOLECULAR BASIS OF PLASTICITY:
There have been many many studies of the basis of critical period plasticity. Hensch's 2005 review that you cite and his 2004 Annual Reviews of Neuroscience review are good sources, but even more has happened since then. You might look up studies by Lamberto Maffei as well as Hensch, Mark Bear, Stryker, Nigel Daw, and I've probably left out many important players. I don't think the neurotrophins are seen as such central actors at this point, but I am not an expert on this.
The main problem here is that you go to great length to write down all the equations but you do not give us any insight into what has been learned from these studies.
For example, I can tell you what I think we learned in the Miller et al. 1989 studies. As you note, we showed that the development could be described in terms of three sets of functions: those describing input correlations, intracortical interactions, and the arborization or spread of connections from a single input point across cortical cells. We then showed the role of each of these functions in determining the development, within the modeling framework we studied. We showed that input correlations determine receptive field structure, such as whether or not ocular dominance segregation develops on individual cells; that intracortical interactions determine the mapping of receptive field features across cortex; and that the arbor function serves as a "window" on both input correlations and intracortical interactions. In particular, the key determinant of whether or not ocular dominance segregation occurs is the function describing the difference between same-eye correlations and opposite-eye correlations, as a function of retinotopic separation. The condition for ocular dominance segregation to occur is that this function should be positive at least at short distances, and not significantly negative within an arbor radius -- that is, correlations should be stronger within an eye than between eyes at least locally, and the reverse should not be true within an arbor radius. (And the 1989 article mentioned, and the 1994 Miller and MacKay article in Neural Computation cited below made clear, that if there were positive correlations between the eyes, then constraints on total synaptic strength had to be implented subtractively; if they were implemented multiplicatively, then the two eyes had to be anticorrelated in order for ocular dominance segregation to occur). When ocular dominance segregation does occur, then it can be organized into a periodic map. If the intracortical interactions drive periodic activity with a wavelength less than an arbor diameter, then this period will be the period of the ocular dominance columns. If it does not drive periodic activity or drives a much longer period, then the period can be equal to the arbor diameter if there is a constraint such that neither eye's projection can lose overall strength.
None of this has been tested very rigorously, but there have been some tests. The Weliky et al. study of early LGN correlations that I mentioned showed that indeed within-eye correlations were stronger than between-eye. There was some controversy for a while as to whether input correlations might affect column width -- there was a preliminary result that strabismus led to wider ocular dominance columns -- but in the end, it was shown that strabismus did not affect column width (Rathjen, Schmidt and Lowel 2002, cited at the beginning of this review). And Hensch and Stryker showed that alterations in the intracortical interactions could indeed alter column width:
Hensch TK, Stryker MP. Columnar architecture sculpted by GABA circuits in developing cat visual cortex. Science. 2004 Mar 12;303(5664):1678-81.
Those authors gave an interpretation of the effects of their changes on the intracortical interaction function under which the changes they saw as predicted by Miller et al. (1989). However, whether that is the correct interpretation remains to be seen.
If you are going to present various models, I think the thing of interest is to figure out what they learned, what are the strong or robust insights into the determinants of development under some scenario, what are the testable predictions. Just listing the equations in great detail is not very helpful.
Finally, I have to add, I find reviewer 1's suggestion that you cite Goodhill 1994 with respect to the effects of multiplicative vs. subtractive normalization a bit odd. I don't really want to promote myself, but I think it's an objective fact that the reference that both introduced and explained these differences was
Miller, K.D. and D.J.C. MacKay (1994). The Role of Constraints in Hebbian Learning, Neural Computation 6, 100-126.
which followed on a footnote in the 1989 Miller et al. paper. (Those early issues of Neural Computation are not indexed in pubmed. You can find the paper on my web site if you can't find it elsewhere.) We studied a linear model in some generality. Goodhill then took the framework from us and applied it to a nonlinear model with two synapses (he clearly acknowledges this in his paper). I don't mind if you cite the Goodhill paper, but it does seem odd to point to that as the primary source.
We thank the reviewer for his thoughtful comments. We have modified the article based on those comments; we add a detailed point-by-point discussion below.
- We have modified the introductory paragraph in order to include additional recent work showing the great variability of OD spacing in different species. We also emphasized the fact that not all species have OD columns.
- In the ‘Development of ocular dominance’ section we have included a discussion about the findings by Crowley and Katz (1999, 2000). We have included in the discussion the results from Weliky et al. (1999 - 2002).
- We have substituted Fig. 4 (from Sengpiel et al.) with another figure that combines it with Fig. 10 from Crair et al. 2001, as suggested by the reviewer.
- Critical period and the role of electrical activity: we removed this section and placed the discussion about early experiments and the onset of ocular dominance segregation in the section ‘Development of ocular dominance’.
- We included a reference to Huberman et al. (2006) (disruption of retinal waves).
- We have shortened the number of equations and added a more detailed discussion of the model by Miller et al. (1989).
- We have modified the citations regarding the effects of multiplicative vs. subtractive normalization. We have cited Miller and MacKay (1994) as the primary source.
REVIEWER 1 COMMENTS ROUND 3
First, aplogies for the delay in doing this. I am almost ready to accept the article which seems much improved but I am still concerned about the difference between the title - 'ocular dominance' and the subject of the article as written, which is 'ocular dominance columns'. I would suggest either changing the title to 'ocular dominance columns' or explaining what ocular dominance is, and mentioning its presence in rodents (as I pointed out in my second set of comments) in the absence of columns. The Hubel & Wiesel 7 point scale is now mentioned briefly I still think it would be useful to add just a bit more and say what the numbers mean (i.e. 4= equal responses to the two eyes, 1= exclusively contralateral, etc.). To complicate things, the wiki entry on ocular dominance is about a nearly completely different use of the term. Perhaps some clarification of this could be added (as should be done for the wiki article I would think).
I would also remove the reference to Swindale (1988) for human odc width.
We thank the reviewer for his suggestions. We have modified the Scholarpedia entry according to these suggestions:
- We suggested the title 'Ocular dominance columns' to the editor.
- We added a brief, but more detailed, description of Hubel-Wiesel 7-point scale.
- We removed the reference to Swindale (1988) for human ocular dominance width.
- We added a brief paragraph at the end of the introduction mentioning the different uses of the term 'ocular dominance' .
______________________________________________________________________________ Reviewer 1 Final comments:
I will press the 'accept' button after writing this. I have just added a couple of questions/suggestions in square brackets and corrected a few minor typos. I have not gone over it again with a fine toothcomb but I think it is much improved and is a very useful accessible review of the topic.